Diseases, Injuries, and Risk Factors in Child and Adolescent Health, 1990 to 2017

Key Points Question How have the levels, trends, and leading causes of child and adolescent mortality and nonfatal health loss changed from 1990 to 2017? Findings This study found that child and adolescent mortality decreased throughout the world from 1990 to 2017, but morbidity has increased as a proportion of total disease burden, because the major causes of nonfatal health loss during childhood and adolescence have not changed dramatically. Meaning As the global health community continues to prioritize child and adolescent health during the Sustainable Development Goal era, careful attention should also be placed on examining and addressing nonfatal illness and disability across the development spectrum.

C hildhood and adolescence are vulnerable periods and a crucial window for adult health determination. While improvements in the mortality rate of children younger than 5 years (the population often called under-5) have been undeniably dramatic and positive, 1 the full story of child and adolescent health is more nuanced and heterogeneous, with a notably broader range of characteristics than can be told with a single summary statistic. 2 The effects of acute and chronic infectious diseases, nutrition, physical functioning, mental health, and intellectual development set the stage for both individual prosperity and the future human capital of all societies. 3 Eleven of the 18 Sustainable Development Goals (SDGs) and 19 of the 53 health-associated SDG indicators are about child and adolescent health. 4,5 These include ending all forms of malnutrition (SDG 2.2), reducing maternal mortality ratio to fewer than 70 per 100 000 live births (SDG 3.1), decreasing neonatal and under-5 mortality rate to fewer than 12 and 25 per 1000 live births, respectively (SDG 3.2), ensuring universal access to reproductive health care (SDG 3.7), and multiple objectives aimed at combating specific causes of health loss, such as malaria, tuberculosis, HIV, road traffic crashes, air pollution, substance abuse, and noncommunicable diseases (NCDs). However, many of the leading drivers of health loss among children and adolescents are notably absent from the SDG agenda. 6 We have compiled this third annual global report to detail the levels, trends, causes, and correlates of health loss from birth through age 19 years. It reflects several notable improvements from Global Burden of Disease (GBD) 2017. First, we have generated a complete set of internally consistent demographics estimates, with uncertainty intervals (UIs), for agespecific fertility, population, and all-cause mortality. 7 Second, 5 additional countries (Ethiopia, Iran, New Zealand, Norway, and Russia) were estimated at the subnational level. Third, in addition to adding many new sources of data, we have improved data-processing algorithms. Methods for redistributing deaths coded to nonspecific, implausible, or intermediate causes of death were updated to incorporate statistical uncertainty of cause reassignment. Clinical administrative data (hospital and claims) processing methods were updated to better account for hospital readmissions, multiple clinical visits, and ordering of discharge codes by age, sex, location, and time. Fourth, we have improved the epidemiological transition analysis through improved estimation of the SDI.

Methods
Comprehensive descriptions of each analytic component of GBD 2017 are detailed elsewhere 1,7-12 and compliant with the Guidelines for Accurate and Transparent Health Estimates Reporting. 13 The GBD 2017 included 11 467 unique sources for cause of death estimation and 26 007 for estimation of nonfatal health loss. Data sources for each cause-level analysis are available online at the Global Health Data Exchange. 14 The GBD 2017 used a geographic hierarchy of 7 superregions (high-income countries; Latin America and the Caribbean; North Africa and the Middle East; South Asia; sub-Saharan Africa [SSA]; Central Asia, Central Europe, and Eastern Europe; and Southeast Asia, East Asia, and Oceania) containing 21 regions and 195 countries and territories. Fifteen countries were estimated at the subnational level: Brazil, China, England, Ethiopia, India, Indonesia, Iran, Kenya, Mexico, New Zealand, Norway, the United States, Russia, Sweden, and South Africa. Estimates were produced for male individuals and female individuals separately in each of 23 standard age groups. We cover the first 7 of these age groups in this report: early neonatal (0-6 days' postbirth age), late neonatal (7-27 days' postbirth age), postneonatal (28-364 days' postbirth age), 1 to 4 years, 5 to 9 years, 10 to 14 years, and 15 to 19 years. Each of 359 diseases and injuries were arranged in a 4-level mutually exclusive and collectively exhaustive cause hierarchy; most were analyzed as causing both death and disability. The first level (level 1) of the cause list has 3 categories: communicable, maternal, neonatal, and nutritional conditions (CMNN); NCDs; and injuries. At level 2, there are 22 cause groups, and level 3 includes more disaggregated causes of burden (169 causes), as does level 4 (293 causes). The full GBD cause list, including corresponding International Classification of Diseases, and Tenth Revision (ICD-10) codes, is detailed in appendices to the GBD 2017 summary publications. 8,9 All-cause mortality, cause-specific mortality, and years of life lost (YLLs) were estimated using standardized approaches of data identification, extraction, and processing to address data challenges such as incompleteness, variation in classification systems and coding practices, and inconsistent age group and sex reporting. Nonfatal estimates were generated using data from literature, hospital discharge and claims data systems, cross-sectional surveys, cohort studies, case notification systems, and disease-specific registries. Causespecific years lived with disability (YLDs) were calculated by multiplying sequela-level prevalence with corresponding disability weights that were derived from population and internet surveys of more than 60 000 persons and adjusted for comorbidity through microsimulation. 15,16 Disability-adjusted life years (DALYs) are the sum of YLDs and YLLs and used to measure the comprehensive health status of a population for a given location, sex, year, and age combination.
We sampled 1000 draws of the posterior distribution of quantity at the most granular level of each analysis, and 95% UIs represent the range of values between the ordinal 25th and 975th draws. Unlike confidence intervals, which only capture sampling error in a single statistical test, UIs also incorporate uncertainty from other associated steps. Aggregate estimates (eg, DALYs, combined age groups, geographical groups) were calculated by summing draw-level results assuming independence of each quantity. All draw-level results were summarized as mean values and 95% UIs.
We performed 3 secondary analyses for this report. First, we decomposed probability of death from birth to 19 years to illustrate how cause-specific trends are associated with overall survival improvements. Second, we explored the historical association between burden metrics and the SDI, a composite indicator of development based on per capita income, adult education, and total fertility rate for individuals younger than 25 years. 1 Each GBD location's SDI can vary by year, but for reporting purposes, each was assigned to an quintile based on its SDI in 2017. A map of SDI quintile assignments is shown in eFigure 1 in the Supplement and SDI values for each country by year are in eTable 1 in the Supplement. Observed values are the actual disease burden rates in each location-year, while expected values were determined by Gaussian process regression on the range of rates observed for each level of SDI. Third, given the intricate association between the health of women and their children, we examined the historical association between maternal mortality and DALY rates of children and adolescents.
We present a number of different formulations of results in the GBD 2017. Total number illustrates the cumulative size of burden, rates best compare between differently sized populations, and cause fraction (%) allows the comparison of relative importance of specific causes. We refer to those younger than 28 days as neonates, those younger than 1 year as infants, those younger than 10 years collectively as children, and those aged 10 to 19 years as adolescents. We focus on presenting aggregate results for the global level, SDI quintiles, and the GBD regions, either for birth to 19 years en bloc or for infants, children, and adolescents separately. Except when noted, results are for both sexes combined. More granular results are publicly available in an interactive online visualization tool called GBD Compare (https://vizhub.healthdata.org/gbdcompare/) and for download from the GBD Results Tool (http:// ghdx.healthdata.org/gbd-results-tool).

All-Cause Mortality and Decomposition of Causes of Death
Premature mortality is the dominant component of health loss in children and adolescents. The Table shows deaths by age group globally and by SDI quintile. eTable 2 in the Supplement shows the same for superregions, regions, countries, and territories. All-cause child and adolescent deaths decreased 51.7% from 13.77 million (95% UI, 13.60-13.93 million) in 1990 to 6.64 million (95% UI, 6.44-6.87 million) in 2017. More than half (60.1% [95% UI, 59.6%-60.5%]) occurred in infants younger than 1 year and, of those, 46.6% (95% UI, 46.0%-47.3%) occurred in the first week of life. The fastest decline was among children aged 1 to 4 years, in whom global deaths decreased by 61% from 3.62 million (95% UI, 3.52-3.72 million) in 1990 to 1.40 million (95% UI, 1.34-1.48 million) in 2017. Over the same period, mortality decreased 51% to 3.99 million (95% UI, 3.85-4.14 million) in infants younger than 1 year, by 52% to 0.41 million (95% UI, 0.40-0.42 million) in children aged 5 to 9 years, and by 27% to 0.84 million (95% UI, 0.82-0.85 million) in children aged 10 to 19 years. Improvements by age were similar across SDI quintiles.
There were a total of 50 countries where the probability of death by self-harm and interpersonal violence increased between 1990 and 2017. Nine of them had increase of more than 0.1% in the overall probability of death owing to self-harm and interpersonal violence between birth and age 20 years: Syria (1990, 0.04%; 2017, 3.45%), Iraq (1990, 0.26%; 2017, 1.22%), Yemen (1990, 0.07%; 2017, 0.89%), Central African Republic  7.0 (6.7-7.3) 7.6 (7.3-7.9) 4.3 (4.1-4.5) 10.0 (9.8-10.2) 12.4 (12.2-12.5) 12.9 (12.7-13.1) 7.8 (7.5-8.1) 6.7 (6.5-6.9) 7.4 (7.2-7.6) 12.7 (12.     Historical patterns in SDI and DALY rates illustrate the epidemiologic transition by age (eFigure 5 in the Supplement). The slope of the SDI gradient decreased with increasing age for all causes. For CMNN causes, DALY rates tracked closely with SDI differences in all regions, except Southern sub-Saharan Africa and the Caribbean. The association between SDI groups and NCDs was similar in all children younger than 10 years (range across all SDI levels: children younger than 1 year, 30 413.3-63 232.9; aged 1-4 years, 3504.3-12 070.2; aged 5-9 years, 3273.1-5017.1), but flattened somewhat in adolescents (range across all SDI levels: aged 10-14 years, 4757.6-5685.7; aged 15-19 years, 6835.6-8537.9). In this group, the DALY rate was also higher. In several regions, as evidenced by steep temporal slope of regional plots, especially in Andean Latin America and East Asia, improvements in NCD DALYs outpaced what would have been expected on the basis of SDI improvements alone (O:E DALY rates: Andean Latin America, 1990America, -20002000East Asia, 1990-20002000.1). A trend toward increasing DALY rates owing to injury in children aged 5 to 9 years and 10 to 19 years was seen at the lower end of the development spectrum. Eastern Europe and Southern sub-Saharan Africa had consistently higher injury DALY rates than expected by SDI grouping.

Identifying Exemplars
Changes in the ratio of observed-to-expected (O:E) DALY rates from 1990 to 2000 and 2000 to 2017 are mapped in Figure 3 (and by age group in eFigure 6 in the Supplement). Before 2000, 117 countries improved more than expected on the basis of SDI changes, while 116 countries did so after 2000. Seventy-six countries had faster than expected improvement in both periods; the O:E ratio of all-cause DALY rates were most notable in Liberia (1990Liberia ( -20002000, −30.0), Niger (1990-20002000, Kyrgyzstan (1990Kyrgyzstan ( -20002000, −21.0), Peru (1990-20002000, −19.0), and Georgia (1990-20002000 DALY rates: 1990DALY rates: -20002000, 99.2), Equatorial Guinea (all-cause O:E DALY rates: 1990-20002000.1), Bosnia andHerzegovina (all-cause O:E DALY rates: 1990-2000, 29.8;2000, 34.2), and Lesotho (all-cause O:E DALY rates: 1990-20002000. Syria was also an outlier in how much worse than expected observed DALY rates were in children and adolescents (O:E ratios in 1990: all-cause, 0.48; injuries, 0.49; O:E ratios in 2017: all-cause, 1.09; injuries, 4.73). This was primarily owing to increased injury rates (for all individuals younger than 20 years: 1990, 189 305 [95 UI, 154 987-224 769] Corresponding maps depicting data for children younger than 1 year, 1 to 4 years, 5 to 9 years, and 10 to 19 years for CMNN, NCDs, and injuries separately are shown in eFigure 7 and eTable 3 in the Supplement. For most countries in sub-Saharan Africa, improvements were much faster than expected between 2000 and 2017 for children aged 1 to 4 years in particular, with several countries also having more rapid DALY improvement than expected in children younger than 1 year and aged 5 to 9 years. Among adolescents, on the other hand, there was little evidence of accelerated improvement after the turn of the century, with almost half of the countries in sub-Saharan Africa lagging behind expected improvements in DALY rates.  Also notable was the burden of sudden infant death syndrome (SIDS) in infants; SIDS was ranked third cause of DALYs in children younger than 1 year in the high-SDI quintile and in the top 10 for all high-income countries, plus all of Eastern Europe and Central Europe, accounting for 0.71% of deaths in the late neonatal period (age range, 7-27 days) and 2.24% in the postneonatal period (age range, 28-364 days) in 2017 globally. Sudden infant death syndrome also accounted for 3.4% of postneonatal deaths and 17.0% of deaths in the late neonatal period in high-SDI locations, but only 0.67% and 2.15% of deaths, respectively, in low-SDI settings.
In this case, there was barely any association between SDI level and the YLL-to-YLD ratio until the highest SDI strata, which may have reflected a poor penetration of prevention and treatment services for congenital birth defects and neoplasms outside of high-income countries. In the case of NCDs, increasing SDI was associated with an increased YLL-to-YLD ratio in children aged 5 to 9 years and 10 to 19 years.
Neonatal disorders was the only level 3 cause that ranked in the top 10 of both mortality and disability globally, ranking among the top 10 causes of YLDs in many countries in North Africa and the Middle East and sub-Saharan Africa. Musculoskeletal and mental health disorders (including anxiety disorders, conduct disorder, depression, autism spectrum disorders, and drug use disorders) were all highly ranked in highincome countries, in central and eastern Europe, and throughout Asia, Latin America, and the Caribbean. Hemoglobinopathies, such as sickle cell disorders and thalassemias, were also in the top 10 by O:E ratio in a number of countries, including Yemen

Associating Maternal Health Outcomes With Those of Children and Adolescents
To evaluate the association between population-level trends in child and adolescent DALYs and those of their mothers, we compared percentage change from 1990 to 2017 in all-cause DALY rates for children younger than 1 year, 1 to 4 years, 5 to 9 years, and 10 to 19 years with percentage change in death rates owing to maternal disorders for women aged 10 to 54 years (eFigure 11 in the Supplement). There were strong correlations between trends in maternal death and all-cause DALY rates in all childhood age groups (<1 year, r = 0.589; 1-4 years, r = 0.452; 5-9 years, r = 0.507; 10-19 years, r = 0.379); those countries with the most improvement in maternal mortality also tended to have higher performance in reducing child and adolescent deaths. Statistical correlation was strongest for children younger than 1 year of age (r = 0.59), but continued even to health outcomes of older children and adolescents (r range = 0.38-0.45). The overall association between trends in maternal mortality and all-cause child and adolescent DALY rates became stronger after 2000 in all SDI quintiles other than high-middle SDI quintile (low: r = 0.539 in 1990-2000vs r = 0.672 in 2000low-middle: r = 0.540 in 1990-2000Global Burden of Diseases, Injuries, and Risk Factors in Child and Adolescent Health, 1990 Original

Discussion
Children and adolescents in every country in the world were more likely to reach their 20th birthday in 2017 than ever before, but progress in improving health outcomes has been uneven. Mortality reductions were most rapid in children between the ages of 1 and 4 years, driven by global declines in deaths owing to diarrhea, lower respiratory infection, and other common infectious diseases. Improvements accelerated after 2000. The largest absolute declines were seen in Western, Eastern, and Central sub-Saharan Africa, while the fastest rates of decline were seen in East Asia, Andean Latin America, and South Asia. The pattern of change was closely associated with gains in sociodemographic development and temporally aligned with increased development assistance for health, which led to broad improvements in vaccination, early childhood nutrition, sanitation, clean water, and targeted interventions for HIV/AIDS and malaria. 3,17-20 A vast unfinished agenda in child and adolescent health remains. While malaria has decreased dramatically across the African continent, there are many countries, especially in western sub-Saharan Africa, where parasite transmission, acute illness, and mortality from malaria remain high. Lower respiratory infection, diarrhea, and acute malnutrition also remain among the top killers of children and adolescents in the world in 2017. Investment in programs targeting prevention and effective syndromic treatment of CMNN disorders clearly pays dividends, and these investments must continue. In locations with higher SDIs, a continuing shift toward nonfatal health loss from NCDs, such as congenital birth defects, mental and behavioral disorders, injuries, and asthma are challenging health systems to adapt. 21 The consistent burden of NCD-attributable DALYs in adolescents over the past 28 years illustrates a need for continued research and action on NCDs as communicable disease burden declines across the development spectrum. The burden of injuries in adolescents surpasses that of CMNN causes throughout the study period for middle-SDI through high-SDI countries, and with the relative faster decline of CMNN causes in low and low-middle countries, the relative ranking of injuries may switch in those locations in the near future.
Overall health improvements were slowest in adolescents. Few locations showed any evidence of improvements in health among adolescents that exceeded the trends expected with general societal development gains. Adolescence is a key phase of the life course and human development, including a phase of growth and maturation of the reproductive, musculoskeletal, neurodevelopmental, endocrine, metabolic, immune, and cardiometabolic systems into adulthood. 22 Gains or lack thereof in adolescent health thus have the potential to influence individual and societal outcomes for periods substantially longer than the teenage years. In terms of family and home life, key issues include the improvement of sanitary and living conditions, stable food systems, quality education, and gainful employment. 23 Also, HIV/ AIDS remains an imminent threat to the health and wellbeing of older children and adolescents in many countries in sub-Saharan Africa, such as South Africa, Zimbabwe, Lesotho, Eswatini, Botswana, and Zambia. The large and growing burden of mental health and substance use disorders among older children and adolescents also is an emerging threat to the thrive component of the SDG survive and thrive agenda. While the psychological needs of children and adolescents show similarities across geographical settings, [24][25][26][27] comparatively little is understood about modifiable risk factors or effective prevention programs for childhood mental illness, outside of ensuring that caregivers are attuned to the link between mental health disorders and self-harm. 28,29 Injuries in general continue to be a major cause of early mortality and longterm disability among older children and adolescents in all countries. While many types of injuries, such as those arising from war and natural disasters, may not be preventable with health sector-based approaches, diligent preparedness planning can help mitigate the immediate health aftermath of them. [30][31][32] Others are much more amenable to policies and programs that focus on prevention using what have come to be regarded as common-sense safety measures, such as speed limits, seat belts, and cycle helmets for road traffic accidents, 33,34 fencing around water hazards and swimming-skills training for drowning, 35 and policies to prevent self-harm via improving safety and limiting access to firearms and chemicals. 36,37 At the other end of the age spectrum, neonatal disorders remain a major prevention and treatment challenge, especially for countries outside the high-SDI quintile that lack the same level of financial and human resources to dedicate to the intensive care needs of sick neonate. Investment is needed to develop and implement cost-effective interventions for neo-natal disorders that take into account the dynamics of maternal health, risk-factor exposures during pregnancy, clinical care systems, supportive equipment needs, and the cultural differences around how families and communities care for newborns. It is important also to invest in the ongoing care of children who survive perinatal emergencies only to develop longterm complications, such as cerebral palsy. Congenital birth defects and hemoglobinopathies are 2 other groups of causes for which there is little evidence of improved outcomes outside the high-SDI quintile, perhaps reflecting the resourceintensive nature of averting deaths owing to such conditions and societal barriers to care 38 but also likely because of a failure of recent clinical advances to be adopted in lowerresource settings. 39 The close linkage between trends in maternal and child health reinforces the notion that the health of different population segments are closely interconnected. 40 The simultaneous focus of the Millennium Development Goals on maternal and child mortality appears to have led to closer association between them since 2000 via alignment of funding streams, targeting of common risk factors between mothers and their children, an increased focus on delaying the age of parenthood by increasing education, contraception, and increased birth spacing, and catalyzing improved gender equity. [41][42][43][44][45][46][47] There are strong ties between the physical health of women (eg, high body mass index, NCDs, nutrition) and neonatal outcomes (such as pregnancy complications, short gestational age, and low birth weight), which are in turn linked with poorer health outcomes and delayed development. 3,6 This is to say nothing of the potential epigenetic connections between mothers and the health of their children that have the potential to extend beyond the neonatal period into childhood, adolescence, adulthood, and the next generation. 48 The subset of countries that are outliers to this trend of concomitant improvement in maternal and child health warrant close examination to determine the underlying causes. Challenges are likely to arise whenever funding streams are decoupled, education or family planning programs are disrupted, or the health of young women is not prioritized.
The epidemiological transition has unique implications for the health of children and adolescents and the potential trajectory of socioeconomic development. In particular, as more children survive, the human capital potential societies will expand, but as more children with health problems are also surviving, there is potential for increased burden on health and education systems. The cost of sustaining progress on child and adolescent health and well-being is not insignificant. To achieve the goal of surviving and thriving and realized the human capital potential of children and adolescents, all countries must make strategic investments in education and health systems, including human resources for health, supply chains, infrastructure, governance, and increased support for children with developmental disabilities. Alignment of funding around interconnected drivers of human development and health loss is also required to achieve the SDGs. 49 The SDGs are expansive, but they should not be considered a comprehensive rubric for achieving improved child and adolescent health. For example, outside of women's repro-ductive health and experiences of sexual violence during adolescence, the SDG goals, targets, and indicators remain largely silent on the unique social, environmental, and biological determinants of health occurring in adolescence across the socioeconomic development spectrum. This blind spot in international health targets, planning, and prevention fails to capture the complex transitions occurring during adolescence in particular. Many additional nonhealth SDG indicators also focus on reducing poverty, expanding education, stabilizing environments, strengthening economies, and reducing overall socioeconomic inequality within each country and throughout the world, all of which are relevant to the health and well-being of young persons.

Limitations
The GBD study is an iterative process and, despite continued methodological advancements and improvements in data, this study has a number of limitations. First, all limitations documented in the elements of the GBD estimation process that allow for YLL, YLD, and DALY estimation will contribute to uncertainty in these summary measures. Second, these summary measures of population health are influenced by data availability. Time lags in the reporting of health information by national authorities and thus subsequent incorporation into the GBD estimation mean that these estimates are based on data that are already out of date. Relatedly, data deficiencies from populations in conflict zones (eg, Syria, Iraq, Yemen, South Sudan, Afghanistan), autonomous subnational regions, and certain nongeographical subpopulations (ie, migrants, refugees, and some indigenous peoples) limit the precision of some of the estimated levels and trends of disease burden. Third, the association between YLLs, YLDs, DALYs, and SDIs, although explanatory, cannot be viewed as causal. Fourth, a nontrivial assumption of the analyses is the independence of the uncertainty calculated for YLLs and YLDs. Because of the link between death and prevalence, a positive correlation probably exists between these uncertainties that are not captured in this analysis. Study limitations specific to child and adolescent health include the comparatively poor quality of cause-ofdeath certification in neonates and infants vs older persons, the relatively broad age categorization of all 1-to-4-year-old children in 1 group, and the limited ability to quantify the magnitude of specific intergenerational, societal, and environmental factors that are ecologically suggested by this study.

Conclusions
Globally, the aggregate health status of children and adolescents improved dramatically between 1990 and 2017, particularly owing to declines in death owing to infectious diseases, but nonfatal health loss has increased in both absolute and relative terms, and the gap between best and worst performers has widened. Continued monitoring of the drivers of child and adolescent health loss is crucial to sustain the progress of the past 26 years in the SDG era. The global community must commit to creating systematic accounting of drivers and consequences of long-lasting negative health outcomes beginning in childhood and the effects of long-term morbidity on health systems and human capital and ensuring that no populations are left behind. Only then will we be able to accelerate progress to 2030 and beyond.