Diseases due to unhealthy environments: an updated estimate of the global burden of disease attributable to environmental determinants of health

1Attribution of the burden of disease to environmental risks highlights the importance of environmental protection for people's health and can inform priority setting for targeted management of environmental determinants. Ten years ago the global burden of disease attributable to the environment was estimated for the first time in a comprehensive, systematic and transparent way. The study concluded that as much as 24% of disability adjusted life years (DALYs) and 23% of deaths were due to modifiable environmental risks.

2The health impacts of specific risk factors have traditionally been assessed separately., A comprehensive account of the consequences of unhealthy environments that are modifiable outlines the full potential of disease prevention that can be achieved by reconsidering the way we shape our environment. Since the last assessment 10 years ago, considerable more evidence has become available which justifies an updated assessment. We present here the methods and results of a new study which updates the previous analysis, by compiling the most recent synthesized and other key evidence on each disease and injury and their links to the environment. We present environmental burden of disease both in terms of environment-attributable mortality and DALYs, a weighted measure of death and disability.

3The aim of the study is to quantify the links between disease or injury and environmental risks using CRAs and alternative methods and to derive an estimate of the environmental disease burden, overall, by region and country. For policy relevance, we deliberately focus on those risks which could be prevented or reduced by feasible interventions which modify the environment. The assessment was completed by a review of effective interventions for each of the investigated diseases.

4Environmental health has been defined as that part of public health that addresses all the physical, chemical and biological determinants of health external to a person, and all the related factors impacting behaviours. Included under environment for the purpose of this study are exposure to pollution and chemicals (e.g. air, water, soil, products), physical exposures (e.g. noise, radiation), the built environment (e.g. housing, land-use, infrastructure), other anthropogenic changes (e.g. climate change, vector breeding places), related behaviours and the work environment. Excluded are life style factors and behaviours which have no or only minor relations with the physical environment such as diet, tobacco or alcohol consumption, environments which cannot reasonably be modified (e.g. wetlands, pollen), or social conditions and unemployment. These risks are further detailed in Supplementary File (A1). The focus is placed on disease which can be prevented, either with almost immediate effect, or with longer term transformations.

5For each of the 133 disease and injury groups, we searched the literature systematically using Pubmed and Google Scholar for population health impacts from environmental risks and effects of interventions addressing those risks. The search strategy included a range of different MeSH (Medical Subject Headings) terms and keywords on each of the diseases or injuries, combined with terms for ‘environment’, ‘occupation’, relevant environmental risks and any of the occupational groups at risk, starting from the year 2004 until 2014. Older literature was taken from the earlier study and major projects of risk assessments were reviewed. Furthermore, the literature and data repositories were screened for documented and publicly available data and information on population health impacts, effects of interventions, exposure-response relationships, transmission pathways and causality. Global estimates of population impacts from environmental risks were completed with national or regional estimates, results of systematic reviews and meta-analyses on disease reduction from interventions or on environmental determinants; and finally by individual studies on interventions and environmental determinants. The focus on evidence of interventions underlines risk reductions that are already feasible, whereas other risk reductions may not yet be feasible or performed at large scale. Only risk factors with an established link of causality to health were further considered.

6The population attributable fraction (PAF) of a risk factor is the proportional reduction in population death or disease that would occur if exposure to this factor was removed or reduced to an achievable, alternative (or counterfactual) exposure distribution. To calculate the PAF of a risk factor to a disease, the following information is needed: (i) the exposure distribution to the risk factor within the population of interest, (ii) the relative risk (RR) linking each level of exposure to the specific disease or injury, and (iii) an alternative (counterfactual) exposure distribution to which environmental risks could be reduced. The counterfactual exposure distributions were based either on evidence from interventions, removal of pathways which have been eliminated elsewhere, or exposures achieved in some populations or areas.

7According to the results of the systematic literature review (see above), four different approaches were used to estimate the fraction of diseases attributable to environmental risks in the following order of priority: (i) CRAs, which generally provide estimates based on the highest levels of evidence and most comprehensive data,– (ii) estimates based on more limited exposure information and/or exposure-risk relationships, (iii) diseases with a transmission pathway dependent on specific modifiable environmental conditions were fully attributed to the environment (such as intestinal nematode infections which require contamination of the environment by human excreta), and (iv) expert surveys.

8In priority, we used systematic global estimates of population impacts from environmental risks (CRA type of assessments).,– These assessments are systematic evaluations of changes in population health resulting from modifying the population distribution of exposure from the current situation as compared to an alternative exposure, in combination with corresponding exposure-risk relationships. In these assessments, exposure is assessed for country populations as much as possible, the extrapolability of exposure-response relationship screened. CRA type of assessments are the method of choice and represent the highest level of evidence for environmental health conditions with a clear, established link between exposure and health outcome, such as exposure to air pollution or inadequate water and sanitation, chemicals or radiation. However, often available data is too limited to perform CRA type of assessments such as for insect vectors of diseases or rodent reservoirs of zoonoses which are more difficult to measure or which show a level of variation that is hard to translate in a disease burden, and alternative methods as specified below needed to be used.

9Information on estimating disease burden from a combination of different risks is given in Supplementary File (A2).

10When sufficient exposure distributions, or exposure-risk estimates or other important information was missing to perform CRA type of assessments, estimates based on more limited epidemiological data were performed, such as for HIV/AIDS, Hepatitis B, other sexually transmitted diseases, suicide and underweight.

11Additional information can be found in the full WHO report on this work. For several diseases, approximate epidemiological estimations were also used to support expert opinion (e.g. unintentional injuries from fires).

12Certain infectious diseases are solely transmitted through pathways which depend on specific modifiable environmental conditions, such as intestinal nematode infections which require contamination of the environment by human excreta. These diseases were fully attributed to the environment on the basis of their transmission pathway.

13When estimates of population impacts from environmental risks were not available or could not be developed in the framework of this study, experts were asked to provide a best estimate of the fraction of the specific disease of the global population attributable to the reasonably modifiable environment, as well as the 95% confidence interval (CI). Experts were selected on the basis of their publications in the area of the disease or the relevant environmental risk factor. They were provided with abstracts of search results of the systematic reviews described earlier, as well as an initial estimate that was based on pooled estimates from the literature. Three or more experts were chosen for each disease or injury. More information on generating PAFs and confidence intervals from the experts' replies is given in Supplementary File (A3).

14Where the body of evidence resulting from the updated literature review did not substantially differ or was unlikely to justify a change in experts' estimation of PAF, the results of the expert survey of the previous study were taken.

15To calculate the fraction of disease attributable to a risk factor for any defined population, compiled or estimated PAFs were multiplied by the corresponding WHO disease statistics, by disease or injury, country, sex and age group, and for deaths and DALYs. Equations are listed in Supplementary File (A4).

16The evidence on effectiveness of interventions was further compiled by disease in order to summarize the main intervention areas.

17Results of environment-attributable deaths and disease burden, the attributable fractions, as well as the respective estimation method are listed in Table 1. The environmental fractions of the burden of selected diseases are shown in Fig. 1. Out of the 133 diseases or injuries, 101 had significant links with the environment, and 92 of them have been at least partially quantified. These 92 were grouped in 66 main disease and injury groups. Of these, global CRAs were available for 20 groups, of which 12 could be exclusively used for those diseases and eight needed to be completed by expert opinion. Eight diseases could be assessed (Table 1) on the basis of more limited epidemiological data, and four further disease PAFs were based on their transmission pathways. The PAFs of the remaining 31 diseases were fully estimated through expert surveys. More than 100 experts provided more than 250 quantitative replies. In terms of estimated environmental disease burden (in DALYs), as much as 56% could be estimated with CRA-type methods (of which 36% with a combination of risk factors), 40% were based on expert surveys (of which 8% in the 2015 round), 3% on estimations using more limited data, and 1% based on transmission pathways (Table 1).

18A description of the underlying evidence and region-specific results for each disease or injury are detailed in the report along with compiled effectiveness of environmental interventions. Based on a summary of the literature review on interventions, we report a mapping of diseases to main strategies for disease reduction through environmental improvements in Table 2, which are further detailed in the full report.

19Environmental risks contributed 23% (95% CI = 13–34%) of the global burden measured in deaths, corresponding to 12.6 million deaths in 2012, and 22% (95% CI = 13–32%) in DALYs. In children under 5 years, as much as 26% of deaths and 25% of DALYs are attributable to the environment.

20Global deaths attributable to the environment are dominated by 8.2 million deaths from noncommunicable diseases, followed by 2.5 million deaths related to infectious, parasitic, neonatal and nutritional diseases, and 2.0 million deaths from injuries. The difference is much less important in terms of disease burden, with 276, 202 and 118 million DALYs attributable to the environment in noncommunicable diseases; infectious, parasitic, neonatal and nutritional diseases; and injuries, respectively. Whereas there are significantly more deaths from noncommunicable diseases, infectious, parasitic, neonatal and nutritional diseases and injuries affect the young to a greater extent and therefore lead to relatively higher losses of DALYs relative to noncommunicable diseases (Fig. 2).

21Figure 2 shows that overall disease burden attributable to the environment (thick grey line) peaks for the very young and for adults aged 50–75 years. These two age groups show important susceptibilities to environmental conditions. Children are mainly affected by communicable diseases. For the age group between 50 and 75 years the contributions of infectious diseases and injuries are still significant, while noncommunicable diseases, in particular cardiovascular diseases due to ambient and household air pollution, become very important. Box 1 highlights the shift from environmental disease burden from communicable to noncommunicable diseases between 2002 and 2012. Infectious, parasitic, neonatal and nutritional: PAF from 31% in 2002 to 20% in 2012Noncommunicable diseases: PAF from 17% in 2002 to 22% in 2012Injuries: PAF from 37% in 2002 to 38% in 2012Overall: PAF from 23.3% in 2002 to 22.7% in 2012

22Infectious, parasitic, neonatal and nutritional: PAF from 31% in 2002 to 20% in 2012Noncommunicable diseases: PAF from 17% in 2002 to 22% in 2012Injuries: PAF from 37% in 2002 to 38% in 2012Overall: PAF from 23.3% in 2002 to 22.7% in 2012

23Infectious, parasitic, neonatal and nutritional: PAF from 31% in 2002 to 20% in 2012

24Noncommunicable diseases: PAF from 17% in 2002 to 22% in 2012

25Injuries: PAF from 37% in 2002 to 38% in 2012

26Overall: PAF from 23.3% in 2002 to 22.7% in 2012

27Age-standardized deaths and DALYs by country are provided in Tables A1 and A2 of the Supplementary File. While the highest burden of environment-attributable disease is still in Sub-Saharan Africa and dominated by infectious, parasitic, neonatal and nutritional disease burden, the per capita deaths from noncommunicable diseases are now higher in most other regions of the world. Figure 3 shows environmentally related deaths per 100 000 population by gross national income (GNI). The size of the bubbles is proportional to country population. There is a reduction of deaths with increasing income up to a GNI of around 25 000. At larger incomes there is no difference in death rates, with most countries having around 50 deaths per 100 000.

28Compared to our estimates for 2002, we see a major shift in the importance of environmental factors in noncommunicable disease aetiology. This is due to (i) the composition of the global disease burden which is now dominated by noncommunicable diseases, (ii) increased evidence on environmental determinants of noncommunicable diseases, and (iii) growing importance of environmental factors that contribute to noncommunicable diseases such as air pollution. As the world population continues to age rapidly, the trend of environmental risks predominantly affecting noncommunicable diseases is expected to become more pronounced.

29One hundred and one out of 133 diseases and injuries were at least partially attributable to manageable environmental factors, as compared to 85 out of 102 in the previous study. In addition, the share of estimates based on the highest evidence level, i.e. using CRA type of approaches, has considerably increased and now reaches 56% (for DALYs), as compared to less than 10% in the previous study. In these high-evidence assessments, exposures are being assessed at country level or higher resolution, such as by age and gender to the extent possible and where appropriate, and the transferability of exposure-risk relationships to other population groups than where assessed are being verified or adjusted. This adds to the comprehensiveness and strength of evidence of the previous report.

30Nevertheless, our numbers show that environmental factors continue to contribute to a large disease burden from communicable diseases in many low and middle income countries. In these countries, environmental risks leading to infectious diseases especially in children, such as household air pollution, unsafe drinking-water and poor sanitation and personal hygiene are still highly prevalent., Furthermore the burden from respiratory and intestinal infections in these countries remains high. At the same time they experience the double burden of communicable and noncommunicable diseases.

31Our results differ from the Global Burden of Disease Study 2013 (GBD 2013) which attributed 12% of global DALYs and 16% of global mortality to environmental risks, mainly because we used a broader scope of the definition of environment and complementary methods of assessment. Those risks comprise unsafe water, sanitation and hygiene; air pollution (ambient particulate matter, ozone and household air pollution); second-hand smoke; lead and residential radon exposure; and occupational risks (NB: here we do not count burden attributable to physical inactivity/low physical activity as also for our analysis we did not quantify the environmental part of the burden from this risk factor). Our analysis covers a broader range of environmental risks adding noise (only included as occupational noise in GBD 2013); various chemicals; risks associated with poor housing, the recreational environment, water resource management, land use and the built environment; other community risks; radiation and climate change. Additionally, we consider more risk-factor disease links. Furthermore, GBD 2013 rated high blood pressure as most important risk factor, causing alone as much as 19% of global deaths and 8% of all DALYs. Some of this burden can however be attributed to environmental factors such as air pollution,, arsenic and lead exposure, occupational risks and environmental noise.

32A large part of this analysis is based on surveys of expert opinion and the uncertainties of such estimates are relatively large. However, experts were provided with the body of evidence that was identified during the systematic searches on the particular disease and its links to the respective environmental risks. We only updated this process when justified by a significant change in evidence. Further uncertainties relate to data limitations and assumptions made in e.g. CRA type of analyses.,– Also key exposures at younger ages, which may result in noncommunicable diseases at older ages could not be adequately captured in this study.

33Certain diseases or environmental risk factors were not included in our analysis, either because there was insufficient evidence and therefore health effects were not quantifiable (e.g. changed, damaged or depleted ecosystems and exposure to endocrine disrupting substances), or because the risk factor(s) caused a relatively small disease burden, or is/are of regional significance but do not feature at a global scale. Environmental risks not readily modifiable, e.g. pollen, were not considered.

34Additional conservative approaches have been chosen for this analysis as compared to the previous one in order to increase methodological rigour. For example, (a) only the main environmental risks were quantified where CRA estimates were available, and (b) the exposures of similar risks were combined before the estimation of health impacts. The environmental disease burden measured in DALYs between 2002 and 2012 is not directly comparable as some of the basic parameters as discounting and age-weighting for DALY estimation changed during this period. Using the same methods, the change would have been greater, as more deaths are now due to noncommunicable diseases, which tend to occur at older ages, and induce fewer years of life lost (and fewer DALYs).

35We have not considered health impacts of social determinants. There is, however, a strong link between the conditions of people's daily lives and environmental risks to health. The lower people's socioeconomic status the more likely they are to be exposed to environmental risks, such as chemicals, air pollution and poor housing, water, sanitation and hygiene. Poor people and communities are therefore likely to benefit most from environmental interventions as they are disproportionally affected by adverse environments.

36In principle, and given the methods and definitions chosen, the attributable burden here equates what can be prevented if the risks were removed. While we currently have solutions for reducing many of the prevailing risks, interventions that are affordable and that could completely eliminate certain risks such as ambient air pollution at a larger scale may require further development. Others, such as use of solid fuels, could be removed with almost immediate effect if the necessary means were made available. Yet for exposures which seem unavoidable in the short term, approaches are being considered which would require certain transformations in the way we currently produce and consume.

37Important calls for action are coming from two main global platforms. One of them was created by the adoption of the SDGs in September 2015. It was significant that the Heads of State gathered at a Special Session of the UN General Assembly did not agree on another agenda or declaration, but made a pledge to ‘the transformation of our earth’. Full adherence to the obligations created by this pledge, even if only moral could result in important improvements on the reduction of environmental risks. The Supplementary File (A5, Table A3) gives further information on SDGs and their links with a healthy environment. The other is climate change. International efforts to reduce our carbon footprint (one such example is the recent Paris Agreement, the first global agreement to reduce climate change) would lead to innovative interventions with positive ramifications to several key environmental factors, including to air pollution, water, chemicals, among others.

38This analysis, which confirms that reducing environmental exposures can greatly improve our health and is critical for attaining the SDGs, has been generated considering a large list of environmental risk factors and risk factor-disease links. For half of those links, CRA types of assessment were available basing the results on solid evidence.

39In conclusion, our results convey good news as we included only those environmental exposures that are amenable to change, meaning that interventions exist for removing a large part of global disease burden. A prerequisite would be a stronger focus on primary prevention placing a healthy environment at the centre of such an effort. This is not a task for ministries of health alone. Tackling environmental risks requires intersectoral collaboration. After nearly 50 years of actively promoting this concept, whether referred to as intersectoral action, breaking down silos or the nexus approach, it remains elusive as ever. The statement ‘intersectoral collaboration: loved by all, funded by no-one’ points to obstacles, mainly vested interests, that have burdened this approach ever since it was included as part of the WHO/UNICEF Alma Ata Declaration on Primary Health Care in 1978. Environmental health, quintessentially intersectoral, has suffered most from this lack of progress. There remain a number of health sector-specific functions (monitoring, surveillance), but for the actual interventions the health sector will have to create the enabling environment for intersectoral action. Investing in environmental interventions pays off for governments; it reduces the transfer of hidden costs from other sectors to the health sector. This new report provides the evidence base for intersectoral action providing the evidence to systematically consider the integration of measures into all policy areas.