Measuring the health-related Sustainable Development Goals in 188 countries: a baseline analysis from the Global Burden of Disease Study 2015

1In September, 2015, the UN General Assembly adopted “Transforming our World: The 2030 Agenda for Sustainable Development”, a resolution outlining a new framework to form the cornerstone of the sustainable development agenda for the period leading up to 2030. This new framework replaced the Millennium Development Goal (MDG) framework that expired in 2015, establishing 17 universal goals and 169 targets referred to as the Sustainable Development Goals (SDGs). The SDGs substantially broaden the development agenda beyond the MDGs and are expected to frame UN member state policies over the next 15 years. To measure progress towards achieving the goals, the UN Statistical Commission created the Inter-Agency and Expert Group on Sustainable Development Goal Indicators (IAEG-SDGs) with a mandate to draft an indicator framework that aligns with the targets. The IAEG-SDGs announced a total of 230 indicators to measure achievement of the 169 targets. Health is a core dimension of the SDGs; goal 3 aims to “ensure healthy lives and promote wellbeing for all at all ages”. Health-related indicators—ie, indicators directly pertaining to health services, health outcomes, and environmental, occupational, behavioural, or metabolic risks with well established causal connections to health—are also present in ten of the other 16 goals., Across these 11 goals, there are 28 health-related targets with a total of 47 health-related indicators.

2Evidence before this studySince the adoption of the Sustainable Development Goals (SDGs) in September, 2015, demand to establish independent, robust avenues for monitoring progress for the SDGs has escalated. However, substantial challenges exist in undertaking comprehensive and comparable assessments of health-related SDG indicators to monitor and guide development agendas and health policy implementation.Added value of this studyThe Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) features more than 1870 collaborators from 124 countries and three territories and provides an independent analytical platform through which levels of health-related SDG indicators can be assessed across geographies and over time in a comparable manner. Drawing from GBD, we provide the measurement of 33 of the 47 health-related SDG indicators and introduce an overall health-related SDG index for 188 countries from 1990 to 2015.Implications of all the available evidenceGBD and its analytical framework allow detailed analyses of country-level performance across health-related SDG indicators and over time. This information can be used to identify high-performing and low-performing countries, inform policy decisions, guide resource allocation, and monitor progress towards the health-related SDGs. The varied historical progress in improving a subset of health-related SDG indicators and rising prevalence of risks such as child overweight underscores the complex health landscape the world faces in the SDG era.

3Evidence before this study

4Since the adoption of the Sustainable Development Goals (SDGs) in September, 2015, demand to establish independent, robust avenues for monitoring progress for the SDGs has escalated. However, substantial challenges exist in undertaking comprehensive and comparable assessments of health-related SDG indicators to monitor and guide development agendas and health policy implementation.

5Added value of this study

6The Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) features more than 1870 collaborators from 124 countries and three territories and provides an independent analytical platform through which levels of health-related SDG indicators can be assessed across geographies and over time in a comparable manner. Drawing from GBD, we provide the measurement of 33 of the 47 health-related SDG indicators and introduce an overall health-related SDG index for 188 countries from 1990 to 2015.

7Implications of all the available evidence

8GBD and its analytical framework allow detailed analyses of country-level performance across health-related SDG indicators and over time. This information can be used to identify high-performing and low-performing countries, inform policy decisions, guide resource allocation, and monitor progress towards the health-related SDGs. The varied historical progress in improving a subset of health-related SDG indicators and rising prevalence of risks such as child overweight underscores the complex health landscape the world faces in the SDG era.

9The SDGs were developed through a highly consultative and iterative process that included multiple meetings with expert groups, civil society, and governments. However, the process of developing the SDGs and the accompanying goals, targets, and indicators has not been without its critics. In both scientific settings and the news media, the common refrain has been that the SDGs are a long list of vague goals that lack clear, realistic, and measurable targets and indicators,,,,,,, and that they are not accompanied by a clear theory of change articulating how the pieces fit together. In view of the potential importance of the SDGs in directing national policies and donor investments, there has also been intense debate about the selection of targets and indicators; despite the lengthy list, some think that the SDGs are missing key areas of development, ranging from prohibition of forced labour to improvement of mental health.,, Concerns have also been expressed about the feasibility of measuring the 230 proposed indicators.,, Indeed, measurement of countries' current status and progress towards meeting the SDG targets will be an enormous task and will require collective action across a range of national and international organisations, both governmental and non-governmental. The difficulties of measurement are also further compounded by persistent problems of data availability, quality, and comparability across a host of indicators., Furthermore, measurement of development indicators is accompanied by a high potential for political entanglement, which can lead to distorted estimates.,,, Independent monitoring of the SDG indicators will be crucial if they are to be used to accurately evaluate progress to ensure accountability and drive national and international development agendas towards meeting the SDGs.,,,,

10Despite these concerns, increasing work has been done in the past decade to generate independent, comparable, valid, and consistent measurements of development indicators.,,,,, To measure progress on the SDGs, these existing efforts will need to be leveraged, particularly those that provide comparable assessments of health outcomes and risks across countries and over time. The Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) is a primary example of such an initiative. GBD is an open, collaborative, independent study to comprehensively measure epidemiological levels and trends of disease and risk factor burden worldwide, with more than 1870 individual collaborators from 124 countries and three territories across the full range of development. GBD uses a highly standardised approach to overcome challenges of inconsistent coding and indicator definitions across countries, missing and conflicting data, and time lags in measurement and estimation. Of the 47 health-related indicators included as part of the SDGs, estimates for 33 indicators are presently included as part of GBD. The GBD study also has several mechanisms to ensure independence, including the GBD Scientific Council that meets regularly to review all methods and major data changes, and the Independent Advisory Committee that meets twice yearly to review GBD progress and provide recommendations for strengthening GBD estimates.

11In this analysis, while acknowledging the continued debate about the structure, selection, and construction of SDG indicators, we used the GBD study to assess the current status of these 33 health-related SDG indicators. With this baseline assessment, we developed and estimated a summary indicator for the health-related SDG indicators and documented historical trends for this summary indicator. With the GBD results, we identified countries with the largest improvements between 1990 and 2015 to inform roadmaps and provide a basis for monitoring the health-related SDG indicators.

12GBD is an annual effort to measure the health of populations at regional, country, and selected subnational levels. GBD produces estimates of mortality and morbidity by cause, age, sex, and country for the period 1990 to the most recent year, reflecting all available data sources adjusted for bias. GBD also measures many health system characteristics, risk factor exposure, and mortality and morbidity attributable to these risks. In addition to providing highly detailed standardised information for many outcomes and risks, various summary measures are also computed, including disability-adjusted life-years (DALYs) and healthy life expectancy. For the present analysis, we used estimates from GBD 2015 to provide a baseline assessment for 188 countries. Further details on GBD 2015, which covers 1990–2015, are available elsewhere.,,,,,

13We defined health-related SDG indicators as indicators for health services, health outcomes, and environmental, occupational, behavioural, and metabolic risks with well established causal connections to health. Many of the 47 health-related SDG indicators selected by the IAEG-SDGs are produced as part of GBD. Table 1 outlines the ten goals, corresponding to 21 health-related targets and 33 health-related indicators included in this present iteration of GBD. This table also outlines the definition of the indicator used in this analysis; detailed descriptions of the estimation methods and data sources are given in the methods appendix pp 10–311. For the 14 health-related indicators that were not included in this analysis, their prospects for measurement in future iterations of GBD are described in table 2.

14Direct outputs of GBD that are health-related SDG indicators include mortality disaggregated by age (under-5 and neonatal) and cause (maternal, cardiovascular disease, cancer, diabetes, chronic respiratory diseases, road injuries, self-harm, unintentional poisonings, exposure to forces of nature, interpersonal violence, and collective violence and legal intervention [ie, deaths due to law enforcement actions, irrespective of their legality]), as well as disease incidence (HIV, malaria, tuberculosis, and hepatitis B) and prevalence (neglected tropical diseases). The GBD comparative risk assessment includes measurement of exposure prevalence included as health-related SDG indicators (under-5 stunting, wasting, and overweight; tobacco smoking; harmful alcohol use; intimate partner violence; unsafe water, sanitation, and hygiene; household air pollution; and ambient particulate matter pollution), as well as deaths or disease burden attributable to risk factors selected as health-related SDG indicators (unsafe water, sanitation, and hygiene; household air pollution and ambient particulate matter pollution; and occupational risks).

15Underlying GBD outputs are a range of additional health determinants that contribute to the estimation of morbidity and mortality, for which data are systematically compiled and estimates are produced. For example, GBD comprehensively analyses data from household surveys on vaccine coverage and combines survey estimates with reported administrative data to produce time series of vaccine coverage for all countries from 1990 to 2015. Estimates of vaccine coverage are then included as predictors of vaccine-preventable morbidity and mortality in GBD. Additional health indicators produced as part of GBD and included as health-related SDG indicators in this analysis are: met need with modern contraception among women of reproductive age, adolescent birth rate, skilled birth attendance coverage, and universal health coverage (UHC) tracer interventions. For UHC tracer interventions, we developed an index based on the geometric mean of the coverage of a set of UHC tracer interventions: met need with modern contraception; antenatal care (one or more visits and four or more visits); skilled birth attendance coverage; in-facility delivery rates; vaccination coverage (three doses of diphtheria–pertussis–tetanus, measles vaccine, and three doses of oral polio vaccine or inactivated polio vaccine); tuberculosis case detection rate; coverage of antiretroviral therapy for populations living with HIV, and coverage of insecticide-treated nets for malaria-endemic countries.

16For selected indicators proposed by the IAEG-SDGs, we made modifications to the definition for clarity or on the basis of the definition used in GBD (table 1). For example, Indicator 2.2.2 proposes a measure of malnutrition that combined prevalence of wasting and overweight among children under age 5 years. As childhood wasting and overweight have very different determinants, we opted to report them separately. For childhood overweight, we report prevalence in children aged 2–4 years, the definition used in GBD based on thresholds set by the International Obesity Task Force.

17Further details on the estimation and data sources used for all indicators, compliant with Guidelines for Accurate and Transparent Health Estimates Reporting (GATHER),, are included in the methods appendix pp 10–311.

18To identify broad patterns and more easily track general progress, we developed an overall health-related SDG index that is a function of the 33 health-related SDG indicators (referred to as the health-related SDG index). We also constructed two related indices: one reflecting the SDG health-related indicators previously included in the MDG monitoring framework (referred to as the MDG index) and one reflecting SDG health-related indicators not included in the MDGs (referred to as the non-MDG index).

19Three broad approaches can be used to create composite measures: normative, preference weighted, and statistical. Normative approaches combine each indicator based on first principles or an over-riding construct such as the contribution of each indicator to overall health. Preference-weighted approaches weight each indicator by expressed or elicited social preferences for the relative importance of different indicators. Statistical approaches seek to reduce a long set of variables or indicators into common components of variance using methods such as principal component analysis or factor analysis. In this case, because the SDGs reflect the collective vision of UN member states, we used a preference-weighted approach, assuming that each SDG target should be treated equally.

20To combine indicators, we adopted methods used to construct the Human Development Index, which include rescaling each indicator on a scale from 0 to 100 and then combining indicators using the geometric mean. The geometric mean allows indicators with very high values to partly compensate for low values on other indicators (referred to as partial substitutability). In the methods appendix pp 312–13, we describe results from alternative index construction methods (ie, principal component analysis; the arithmetic mean across targets referred to as complete substitutability; and the minimum value across targets referred to as zero substitutability). Quantitative targets for each of the health-related SDG indicators are not universally specified. As a result, we rescaled each health-related SDG indicator on a scale from 0 to 100, with 0 being the lowest (worst) value observed and 100 being the highest (best) value observed over the time period 1990–2015. We log-transformed mortality and morbidity before rescaling. We then estimated the health-related SDG index by first computing the geometric mean of each rescaled health-related SDG indicator for a given target, followed by the geometric mean of resulting values across all SDG targets. To avoid problems with indicator values close to 0, when computing indices we applied a floor of one to all indicators. This analytic approach weights each of the health-related SDG targets equally. In addition to the health-related SDG index, we also used the same methods to construct an index that represents 14 health-related SDG indicators that were previously MDG indicators and an index representing 19 non-MDG indicators (table 1). Uncertainty in the indicator and indices values was computed using a simulation analysis.

21As part of GBD 2015, we assessed cause-specific disease burden and risk exposure along the development spectrum, providing context on expected changes as countries progress to higher levels of income per person, higher educational attainment, and lower fertility.,,, We conducted a similar analysis by examining the relations of the overall health-related SDG index and each of the individual health-related SDG indicators with the Socio-demographic Index (SDI), a summary measure of development that uses lag-distributed income per person, average educational attainment in the population over age 15 years, and the total fertility rate. The SDI was constructed using the same method for the Human Development Index and the health-related SDG index. Each of the three components was first rescaled on a 0–1 scale, with 0 being the lowest (worst) value observed in the time period 1980–2015 and 1 being the highest (best) value observed. SDI was then computed as the geometric mean of these three rescaled components. To capture average relations, we used a spline regression (ie, piecewise linear regression with so-called knots specifying the intersection between pieces) of the health-related SDG indicators and health-related SDG index on SDI using the full set of data by country from 1990 to 2015. We also compared the health-related SDG indicators with the GBD 2015 estimates of healthy life expectancy to explore the relation between the SDGs and overall health achievement for each country.

22The 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.

23Of the 33 health-related SDG indicators, 21 were associated with a defined target, with 18 of them having an absolute level and three having a target relative to 2015 levels (table 3). The proportion of countries already meeting targets linked to health-related SDG indicators in 2015, as specified by absolute levels to be achieved, ranged from more than 60% for two indicators (maternal mortality ratio and under-5 mortality) to 0% for nine indicators. For these nine indicators, all targets involved full elimination of diseases (eg, tuberculosis, HIV, and neglected tropical diseases), reducing prevalence of health outcomes or risk to 0% (eg, childhood overweight and intimate partner violence), or reaching 100% for intervention coverage or health service provision (eg, skilled birth attendance, met need with modern contraception, and UHC tracer interventions).

24In 2015, the median health-related SDG index was 59·3 (95% uncertainty interval [UI] 56·8–61·8) across all 188 countries. This index was highest in Iceland (85·5, 84·2–86·5), Singapore (85·3, 84·1–86·3), and Sweden (85·3, 84·2–86·2) and lowest in the Central African Republic (20·4, 15·9–24·9), Somalia (21·6, 16·0–25·9), and South Sudan (22·5, 15·5–26·6; figure 1). Differences in the 95% UI range stem largely from differences in the availability and quality of underlying data sources for estimating individual indicators; for example, data were sparser for Somalia than they were for Sweden. Some patterns emerged contrary to what might have been expected. For example, the USA (74·9, 73·6–75·9) ranked 28th, driven by poorer performance on MDG indicators (eg, maternal mortality ratio) than other high-income countries and worse performance on non-MDG indicators—most notably, alcohol consumption, childhood overweight, and mortality due to interpersonal violence, self-harm, and unintentional poisoning. India (41·7, 39·7–43·7), despite rapid economic growth, was ranked 143rd, just below Comoros and Ghana.

25Levels of the health-related SDG index were highly clustered (figure 2), with countries in the highest quintile (≥71·5) located mainly in western Europe, high-income North America, parts of Asia (Japan, South Korea, Singapore, and Brunei), and Australasia. The second highest quintile (62·5–71·5) included countries in southern Latin America, parts of eastern Europe, most of the Caribbean, and a subset of countries across other regions (eg, Mexico, Jordan, Azerbaijan, Malaysia, and Costa Rica), whereas countries in the middle quintile (55·7–62·5) were primarily located in South America; parts of east, central, and southeast Asia; and parts of North Africa and the Middle East. The countries in the fourth quintile (37·8–55·7) were mainly found in south and southeast Asia, southern sub-Saharan Africa, parts of North Africa and the Middle East, and parts of eastern Europe. Countries in western, eastern, and central sub-Saharan Africa, as well as a subset of other countries (eg, Afghanistan, Papua New Guinea, Yemen, and Nepal), dominated the lowest quintile (<37·8) of the health-related SDG index. Although the MDG index was correlated with the non-MDG index, country-level performance on these two indices varied considerably (figure 3). Performing well on the health-related MDG index did not guarantee good performance on the health-related non-MDG index. For example, the health-related MDG index in 2015 was similar for Indonesia (52·3, 49·8–54·6) and South Africa (48·9, 46·0–51·3), but Indonesia had a much higher non-MDG index (64·1, 62·0–66·6) than that of South Africa (42·9, 40·3–45·5). This difference for the non-MDG index was primarily driven by South Africa's lower performance for indicators such as childhood overweight, harmful alcohol use, and mortality due to self-harm and interpersonal violence.

26SDI was highly predictive of the overall health-related SDG index (r2=0·88) and MDG index (r2=0·92; figure 4). The non-MDG index was less well predicted by SDI (r2=0·79). This finding is reflective of the variable relations between individual health-related SDG indicators and SDI (results appendix pp 346–47). For instance, SDI was a poor predictor of mortality due to exposure to forces of nature, self-harm, interpersonal violence, and war (collective violence and legal intervention), as well as childhood overweight, intimate partner violence, and ambient particulate matter pollution. By contrast, SDI was highly predictive of maternal mortality ratio, under-5 mortality, and neonatal mortality, as well as mortality attributable to unsafe water, sanitation, and hygiene. Notably, the overall health-related SDG index also had a strong relation with healthy life expectancy (r2=0·86), a summary measure of population health.

27By subtracting expected levels for the health-related SDG index, on the basis of SDI alone, from observed levels (figure 5), we could identify potential geographical deviations well above or below expected values on the health-related SDG index. Countries that represent substantial deviations from the average might warrant further investigation to understand how and why they are underperforming or overperforming relative to the average. This deviation might be due, for example, to more or less efficient use of resources to improve health. Many countries in western Europe, Latin America, and parts of east and southeast Asia, as well as other countries such as Australia, recorded health-related SDG index levels that were higher than expected on the basis of SDI alone. Many of the countries with a health-related SDG index below expected levels on the basis of SDI were located in southern and central sub-Saharan Africa, eastern Europe and central Asia (eg, Belarus and Ukraine), North Africa and the Middle East, south Asia, and selected countries such as the USA.

28To provide a preliminary indication of potential trajectories in the next 15 years, we assessed absolute changes in the past 15 years for each of the 33 health-related SDG indicators and three summary indices (overall health-related SDG index, health-related MDG index, and non-MDG index). Overall, health-related SDG indicators largely improved since 2000, as summarised by the health-related SDG index; notably, gains in the health-related MDG index generally exceeded improvements in the non-MDG index (figure 6). Across countries, the most pronounced progress occurred for UHC tracer interventions, met need with modern contraception, hygiene, under-5 mortality, and neonatal mortality. Such striking gains for the indicator on UHC tracer interventions reflected the scale-up of antiretroviral therapy and coverage of insecticide-treated nets in malaria-endemic countries since the early 2000s.,, Of note, the relatively small improvement for the indicator on malaria incidence represents the large number of malaria-free countries in both 2000 and 2015. Health-related indicators covered by Target 3.3—which aims to end the epidemics of HIV, tuberculosis, malaria, and neglected tropical diseases, and to “combat hepatitis” by 2030—generally saw moderate progress (median absolute change of 2·7 [IQR −0·2 to 4·6] for HIV incidence and 3·9 [IQR 1·7 to 5·7] for tuberculosis incidence), although minimal changes occurred for hepatitis B incidence (−0·2 [–0·4 to −0·05]). In combination, these trends highlight the need for accelerated progress in order to meet Target 3.3. Substantial improvements occurred for childhood stunting (8·2 [3·5 to 14·2]) and, to a more modest extent, wasting (2·7 [0·0 to 6·0]), yet childhood overweight considerably worsened in the past 15 years (−4·5 [–9·2 to −0·7]). This trend occurred across SDI quintiles, emphasising the need for concerted policy attention to reverse this trend. Alcohol consumption worsened slightly in the past 15 years as well (−0·4 [–2·3 to 0·7]).

29Between 2000 and 2015, distinct patterns for absolute changes in health-related SDG indicators surfaced across SDI quintiles (figure 6). While the indicator for UHC tracer interventions improved across all SDI quintiles, the most pronounced gains occurred in low-SDI and low-middle-SDI countries. Childhood stunting and wasting also improved at a faster pace for the low-SDI quintile than for other quintiles. Notably, mortality measures from the MDG agenda—maternal mortality ratio, under-5 mortality, and neonatal mortality—progressed at a similar pace across SDI quintiles. By contrast, mortality due to road injuries, non-communicable diseases (NCDs), and interpersonal violence declined faster in the higher-SDI quintiles than in the lower-SDI quintiles. Prevalence of smoking also had the largest reductions in countries in the high-SDI quintile.

30Between 2000 and 2015, progress on the health-related SDG index (figure 7), as well as on individual health-related SDG indicators and on the MDG and non-MDG indices (results appendix p 6), was highly heterogeneous across geographies. Since 2000, the largest absolute improvements in the health-related SDG index occurred in Timor-Leste (18·5, 95% UI 16·2–20·8), Bhutan (16·2, 13·6–18·7), and Colombia (15·6, 14·1–16·8), whereas three countries—Libya, Syria, and Chile—experienced significant declines. Declines for the next two countries (Brunei and South Sudan) were between 0 and −0·5 and rounded to 0 in figure 7. Countries with the most pronounced gains for the health-related SDG index were found mainly in east, southeast, and central Asia, as well as parts of Latin America (eg, Venezuela and Honduras). Several countries in sub-Saharan Africa also recorded considerable gains in the health-related SDG index, including Rwanda, Ethiopia, Ghana, Namibia, and Angola.

31To demonstrate the usefulness of these estimates for informing progress towards the SDGs, we also identified the geographies with the largest improvement in overall health-related SDG index between 2000 and 2015, stratified by SDI quintile classification in 2000. The five geographies were: Timor-Leste in the low-SDI quintile, Tajikistan in the low-middle-SDI quintile, Colombia in the middle-SDI quintile, Taiwan (province of China) in the middle-high-SDI quintile, and Iceland in the high-SDI quintile. Based on their gains for the health-related SDG index, these geographies could serve as case studies for understanding potential drivers of progress on the SDGs.

32In Timor-Leste, changes in the health-related SDG index were largely driven by improvements in UHC tracer interventions, skilled birth attendance, met need with modern contraception, under-5 and neonatal mortality, childhood stunting, risk exposure to unsafe water and sanitation, and mortality from war or conflict. This overall improvement was despite worsening measures for childhood overweight, smoking prevalence, and alcohol use since 2000. Tajikistan recorded sizeable improvements across various health-related SDG indicators. Among indicators related to the MDGs, these included both measures of child mortality, childhood stunting, coverage of UHC tracer interventions, malaria incidence, and exposure to household air pollution. Improvements were also noted in mortality due to NCDs, interpersonal violence, and war or conflict, as well as mortality attributable to unsafe water, sanitation, and hygiene and to air pollution. However, several indicators either had minimal progress or worsened in Tajikistan, particularly childhood overweight and intimate partner violence. Colombia's most pronounced improvements since 2000 occurred for many of the non-MDG indicators, which included smoking prevalence and mortality rates due to NCDs, road injuries, interpersonal violence, and war. Sizeable improvements were also recorded for a subset of health-related MDG indicators—namely, coverage of UHC tracer interventions, adolescent birth rates, met need with modern contraception, and unsafe sanitation. Nonetheless, similar to other countries, Colombia had minimal progress in or worsened levels of alcohol consumption and hepatitis B incidence. In Taiwan, marked gains occurred for a subset of health-related SDG indicators previously associated with the MDG agenda (eg, adolescent birth rates and coverage of UHC tracer interventions); in parallel, Taiwan had considerable improvements for many non-MDG indicators, such as smoking prevalence and mortality due to NCDs, interpersonal violence, and road injuries. However, HIV and hepatitis B incidence worsened in Taiwan since 2000, and minimal progress occurred for ambient particulate matter pollution and several maternal and child health indicators. For Iceland, its progress on the health-related SDG health index was primarily driven by improvements in mortality due to NCDs and road injuries, smoking prevalence, adolescent birth rates, and both measures of child mortality. Similar to other countries, particularly those in the high-middle-SDI and high-SDI quintiles, Iceland had little progress in childhood overweight and worsening levels of alcohol consumption.

33Further results are provided in the results appendix, and dynamic visualisations are available online.

34The ambitious SDG agenda is accompanied by numerous goals, targets, and indicators for tracking progress. Leading up to and following the UN SDG resolution in September, 2015, considerable debate surrounded the selection of indicators, including scepticism about the feasibility of their measurement., In this study, we produced independent, highly standardised, and comparable estimates of 33 of the 47 health-related SDG indicators across 188 countries. To facilitate overall tracking, we also distilled these 33 health-related indicators into a health-related SDG index. Our findings show the wide range in this health-related SDG index in 2015, from 20·4 in Central African Republic to 85·5 in Iceland. Our historical analysis of these indicators also shows that progress can be achieved. Notable improvements were recorded for several health-related SDG indicators, particularly those that were also MDG indicators, such as under-5 mortality, met need with modern contraception, and childhood stunting. An index of the 14 MDG indicators that were included in the health-related SDG indicators had a median absolute change of 10·0 from 2000 to 2015, and larger reductions were generally found for countries at the lower end of the development spectrum. Our analysis also highlights the challenges associated with the expanded scope of the SDGs, with several of the non-MDG indicators having minimal improvements (eg, hepatitis B incidence) or worsening (eg, childhood overweight) between 2000 and 2015. This finding is further supported by the highly variable relation between the health-related MDG index and the health-related non-MDG index—good performance on the MDG index did not guarantee good performance on the non-MDG index. The overall health-related SDG index was well predicted by SDI; however, SDI was a variable predictor of the performance of individual health-related SDG indicators, particularly indicators that were not in the MDG agenda. Drawing from GBD, these findings provide a strong, comparable basis for monitoring the SDGs; furthermore, the independent nature of these results can enable accountability mechanisms for the multiple national and international, governmental, and non-governmental actors that must achieve progress on the SDGs.

35These estimates also allow the identification of places that have made substantial progress on the health-related SDG indicators. These findings stand to strengthen the global evidence base of lessons learned for accelerating improvements in the health-related SDGs. The five geographies with the greatest improvement in the health-related SDG index between 2000 and 2015, stratified by SDI quintiles (Timor-Leste, Tajikistan, Colombia, Taiwan, and Iceland), have implemented a range of policies and interventions that might have contributed to their progress.

36For instance, following acute conflict and violence during the late 1990s, Timor-Leste, in concert with the World Bank and other development partners, implemented a series of health sector rehabilitation and development projects in 2000 and 2001 to re-establish the country's health system and improve health service delivery to the poor., In more recent years, health-care reform and financing have topped policy agendas in Timor-Leste, including the Ministry of Health's roll-out of a Basic Health Services Package and Hospital Services Package in 2007 under the pursuit of achieving UHC. Following almost a decade of civil conflict that severely disrupted health service provision, Tajikistan launched a series of health reforms beginning in the late 1990s and introduced a new benefits package for guaranteed health services in 2007. Moreover, after the civil war, the Tajik Government refocused policy attention for initiatives on particular diseases such as malaria; indeed, the country's multipronged malaria programme, which emphasises strong surveillance and cross-border activities with Afghanistan, has now brought Tajikistan close to eliminating the disease. Colombia, which experienced ongoing conflict and violence from the late 1980s to 2003, is globally recognised for its expansion of health insurance and services, especially to the poor. While Colombia's health system reforms began well before 2000 (the country approved its universal health insurance scheme in 1993), coverage increased substantially over time, as have the types of services covered by its insurance scheme (eg, cancer care). During the mid-1990s, Taiwan adopted a universal health insurance system, which is viewed as one of its most successful public entities. Taiwan also enacted many road safety laws between the mid-1990s and early 2000s, including mandatory helmet laws for motorcyclists in 1996 and an extension of seat belt laws to general roads in 2001. Iceland's tobacco control policies have been viewed as some of the world's most comprehensive, and the country's long-standing publicly funded health system provides UHC, a factor that might have contributed to its declines in NCD mortality.

37Such progress also highlights important interactions between development goals and wider contextual factors, such as post-conflict experiences in Timor-Leste, Tajikistan, and Colombia, as well as a rebound in health following the dissolution of Soviet rule for Tajikistan. Furthermore, Taiwan's gains underscore the interplay between advancing economic development and deliberate investments in improving UHC. These vignettes highlight only a fraction of the possible learning for informing action towards improving the health-related SDGs. An important future area of work will be to understand in detail how these and other high-performing geographies have achieved substantial improvements in key SDG indicators.

38Besides showing the feasibility and value of measuring many of the health-related SDGs, our findings also affirm concerns voiced during the SDG development process and following the UN resolution. One criticism of the SDGs was the incredibly ambitious nature of some of the targets, such as Target 3.3, which calls for the end of the epidemics of HIV, tuberculosis, and malaria by 2030. Our analysis of these indicators in the past 15 years suggests that a substantial change in the present trajectory of HIV and tuberculosis incidence will be needed to meet this target, and major technological leaps coupled with universal delivery are likely to be necessary. The vague nature of many of the SDG targets has also been a common criticism;,,, Of the 33 health-related SDG indicators in our study, we identified specific targets for only 21 of them. The absence of specific and attainable targets for SDG indicators, health related or otherwise, undermines the usefulness of the SDGs in driving development agendas, a limitation that can and should be addressed at this early stage of the SDG period.

39Our analysis also represents a step towards producing a more cohesive understanding of the interactions between different SDG goals, targets, and indicators—a widely noted criticism., We show the potential for quantifying these interactions by comparing the relations between education, income, and fertility—components of the SDI—and the 33 health-related SDG indicators and accompanying health-related SDG, MDG, and non-MDG indices. Although we acknowledge the ecological nature of this analysis and its usual caveats, SDI was a strong predictor of the overall health-related SDG and MDG indices, highlighting the general importance of income, education, and fertility, as well as intersectoral action for health-related development. However, SDI was a weaker predictor of the non-MDG index, particularly for indicators such as violence (intimate partner, interpersonal, and collective violence) and ambient particulate matter pollution. This finding shows that a sole focus on increasing income and education and decreasing fertility is probably insufficient to meet the SDG targets. It also raises questions about whether other common drivers, analogous to income, education, and fertility, can be determined and their relations with SDG indicators assessed. Combining this initial assessment of the 33 health-related SDG indicators with an expanded quantification and analysis of other potential drivers is an important future area of work that could help to create a more concise, cohesive, and actionable monitoring framework for the SDGs.

40In this report, we focused on measuring indicators proposed by the IAEG-SDGs. In future years, we plan to address three related sets of issues: first, improved assessment of the health-related SDG indicators measured at present; second, inclusion of the 14 currently excluded health-related indicators in the annualised GBD study; and third, potential expansion of indicators consistent with the framing of the targets. We address each of these issues in turn.

41With the present analysis, we made several modifications that we believe improve several health-related indicators for the purposes of measuring progress towards each health-related SDG target. Future iterations are likely to incorporate further modifications to these and other indicators. First, as noted in the Methods section, rather than reporting on the combined prevalence of childhood wasting and overweight, we assessed and measured them separately. Our results support this decision, since they had divergent relations over time (ie, childhood wasting improved for most countries, whereas the prevalence of childhood overweight generally increased) and with SDI (ie, childhood wasting and overweight were negatively and positively correlated with SDI, respectively).

42Second, the IAEG-SDGs' proposed indicator for harmful use of alcohol is the average national-level consumption per person in litres of pure alcohol. The health and non-health risks associated with harmful alcohol use are a function of not only average consumption at the population level but also use patterns (ie, amount consumed at a given time and frequency of consumption). For this analysis, we reported on the summary exposure value of harmful alcohol use, which takes into account the distribution of consumption and the prevalence of binge drinking.

43Third, we made two modifications to the measurement of disaster (Indicators 1.5.1, 11.5.1, and 13.1.2). For mortality attributable to disasters, we chose to report on the lagged 5 year average of disaster mortality. One of the corresponding health-related SDG targets (Target 1.5) is to “build the resilience of the poor and those in vulnerable situations and reduce their exposure and vulnerability to climate-related extreme events and other economic, social and environmental shocks and disasters”. Focusing solely on the observed mortality caused by natural disaster ignores the role of chance in the occurrence of natural disasters; for example, nations with weak resilience to natural disasters might not experience a natural disaster during a given period of time, whereas those with strong resilience might encounter them more frequently. Taking the moving average of disaster mortality lessens the contribution of chance in assessing progress for this indicator. Nonetheless, this adjustment cannot account for background risk of natural disasters, which varies by geography, and future research efforts could include developing a risk-standardised version of the indicator. For natural disasters, data on missing people and people affected by disaster are not readily available across countries and over time.

44Fourth, for occupational health (Indicator 8.8.1), we reported on age-standardised all-cause DALY rate attributable to occupational risks rather than the suggested indicator, which is limited to fatal and non-fatal occupational injuries. This revision captures a wider set of occupational risks instead of only those that result in injuries, which supports the stated target of promotion of “safe and secure working environments for all workers” (Target 8.8). Fifth, for tobacco use prevalence (Indicator 3.a.1), GBD does not presently assess smokeless tobacco use. Furthermore, smokeless tobacco use has a notably different risk profile to smoked tobacco use, and thus it might warrant a subindicator akin to childhood malnutrition. Sixth, for clean fuels and technology (Indicator 7.1.2), we presently use a more limited definition that covers fuels used primarily for household cooking. Seventh, for homicide (Indicator 16.1.1), GBD does not measure this indicator by displacement or migratory status, and similarly for conflict-related deaths (Indicator 16.1.2), we do not measure deaths by displacement status or by more specified causes.

45Eighth, data gaps also account for limitations in the estimation of the UHC tracer indicator (Indicator 3.8.1). We used a set of tracer interventions that were restricted to reproductive, maternal, and child health, as well as a subset of infectious diseases. There is a paucity of data for the coverage of NCD interventions in particular,,, and for higher-level care. Furthermore, the UHC tracer indicator only captures the use of interventions and not the quality of the intervention provided., As more data become available on the delivery of interventions for NCDs and the modification of key risk factors, this information will be incorporated into revisions of the UHC tracer indicator. Finally, data gaps mean that, in this initial assessment, we have also not been able to include a measure of financial risk protection. Substantial investments are needed in this area to address data gaps to be able to track the central role of health system delivery in improving health.

46Of the 14 health-related SDG indicators that were not included in this analysis, there are several that the GBD does not currently measure but that could be assessed in the future through GBD (table 2). These indicators include the coverage of treatment interventions for substance use disorders (Indicator 3.5.1), which would leverage the work on quantifying incidence, prevalence, and mortality for these conditions. Estimating the proportion of women and girls aged 15 years and older who are subjected to sexual violence by people other than an intimate partner (Indicator 5.2.2) would leverage work already undertaken by GBD on measuring prevalence of intimate partner violence. As part of GBD, we have also assembled a host of population-level data that would facilitate measurement of the coverage of health insurance or public health systems (Indicator 3.8.2), health worker density and distribution (Indicator 3.c.1), and completeness of death registration (Indicator 17.19.2). Other indicators are more difficult to measure because of data gaps or unclear definitions. For example, data sources to measure the proportion of people that feel safe walking alone around the area they live (Indicator 16.1.4) are not readily available for most countries.

47Various commentaries have pointed out the absence of indicators for key health outcomes and determinants. Proponents have argued for indicators for mental health that go beyond substance abuse disorders and suicide;,, other NCDs beyond cardiovascular diseases, cancer, diabetes, and chronic respiratory diseases; diseases related to ageing, including osteoarthritis and Alzheimer's disease;, non-fatal disorders that lead to substantial morbidity (eg, sensory disorders); and a host of major risk factors. Another example is Target 3.3, which aims to combat hepatitis—the indicator only tracks hepatitis B, although the data for hepatitis C monitoring are as robust as those for hepatitis B and a highly effective cure for hepatitis C is available. As shown in this report, our GBD collaboration provides the basis for measuring many of these indicators. The danger is that an exhaustive laundry list of indicators, a criticism already levelled at the present SDG list, would dilute the value of the SDGs in focusing attention on where it is most needed.

48There are several important similarities and differences between our assessment of the health-related SDGs and those produced by WHO and the Sustainable Development Solutions Network (SDSN). Like WHO, we focused on the health-related SDG indicators and did not cover indicators across all goals as SDSN does. With our focus on health, we covered 33 health-related indicators, compared with 21 by SDSN and 32 by WHO. Similar to SDSN, we produced a summary measure for the health-related SDG indicators included in the analysis. Most importantly, GBD uses standardised and internally consistent approaches to generate estimates across causes, risk factors, and underlying indicators. For example, we constrain the aggregation of cause-specific deaths to equal all-cause deaths. Furthermore, GBD also produces a complete set of estimates for 188 countries and for individual years from 1990 to 2015. By contrast, WHO and SDSN draw on disparate sources and methods for estimation and, as a result, report on an incomplete set of estimates by country. SDSN provides estimates for 149 countries, whereas estimates for health-related SDG indicators produced by WHO range from 194 countries for under-5 mortality and neonatal mortality to 109 countries for HIV incidence. WHO and SDSN also do not generate estimates for a consistent set of years. WHO reports 2015 estimates for only seven indicators and combines data from a range of years for ten indicators; for example, WHO combined skilled birth attendance estimates by country ranging from 2006 to 2014. SDSN reports 2015 estimates for only four indicators and combines data from a range of years for eight indicators; for example, country estimates for smoking prevalence from the SDSN report range from 2006 to 2013. Complete, consistent, comparable, and contemporary estimates of health-related SDG indicators are necessary to properly track progress on the SDGs.

49This study has several limitations in addition to the ones we already described. First, all the limitations of GBD relevant to the 33 indicators used here apply.,,, Second, we tried to summarise the complexity of the 33 indicators using a summary measure for the health-related SDGs. Many approaches are available for developing summary measures. Since the SDGs are the outcome of a political consensus building process, we opted to use the stated targets as preferences of UN member states that have agreed to the SDG declaration. Our sensitivity analysis shows that using alternative weighting schemes produces broadly similar results (methods appendix pp 312–13). Our sensitivity analysis also highlights the limitation of the statistical approach (ie, principal component analysis) for constructing an index for this purpose, with the first principal component including both positive correlations with indicators such as maternal mortality ratio and negative correlations with indicators such as alcohol use. An alternative could be to weight each indicator by their contribution to healthy life expectancy. Third, we opted to construct the summary measure using the Human Development Index method of rescaling each component on a scale of 0 to 100, and then taking the geometric mean of the components. We chose to use the minimum and maximum observed values to rescale, as targets for all indicators are not clearly specified; however, the limitation of this approach is that minimum and maximum values might change in the future. In the next iteration of this analysis, we will use targets for all indicators and rescale them accordingly; to establish targets for indicators currently lacking explicit ones, we will determine plausible targets based on forecasts of trends through 2030. Fourth, a clear limitation, as highlighted by the UHC tracer indicator, is the need for broad investment in data systems in countries to properly assess progress on key health and development indicators such as the SDGs. As an example, there remains considerable uncertainty about levels and age patterns of mortality and the cause of death structure. Investments in high-quality vital registration systems and other related data collection systems, from censuses and household surveys to health management information systems, are crucial to the proper monitoring of progress towards the SDGs.

50Our GBD collaboration aims to address several of the limitations noted above in future reporting of the health-related SDGs on an annual cycle. As noted, we will also leverage work that is underway to forecast country-specific disease burden, which will additionally provide information on the future trajectory of health-related SDG indicators based on historical trends and provide an explicit way to understand how those trajectories could be changed with different policy adoption. We will also address, in a staged manner, the absence of measures of geographical and socioeconomic inequality in the health-related SDG indicators.

51The measurement of 33 health-related SDG indicators presented here is the product of an extensive, open collaboration that represents many countries across a broad range of development. We invite others to join in this effort to produce an independent, robust basis for monitoring and assessing progress towards the SDGs. Independent measurement is a crucial component of accountability, but it is not the only component. These results should ideally be used as the basis for review and action at the country level. We hope that this collaboration is a major contribution to creating a culture of accountability for the SDGs. Other actors, especially governments, civil society organisations, donors, and global development institutions, need to participate in the process of using this information to enhance accountability through open and transparent review and action.

52Correspondence to: Prof Christopher J L Murray, University of Washington, Institute for Health Metrics and Evaluation, 2301 5th Avenue, Suite 600, Seattle, WA 98121, USA cjlm@uw.edu

53Correspondence to: Prof Christopher J L Murray, University of Washington, Institute for Health Metrics and Evaluation, 2301 5th Avenue, Suite 600, Seattle, WA 98121, USA cjlm@uw.edu

54See Online for infographic http://www.thelancet.com/infographics/SDG

55See Online for infographic http://www.thelancet.com/infographics/SDG