Chapter 8: Estimating the Burden of Disease from Climate Change Protecting our Health from Climate Change: a Training Course for Public Health Professionals Chapter 8: Estimating the Burden of Disease from Climate Change

Overview: This Module Outlines steps involved in estimating the burden of disease from climate change Presents worked examples for several of the health impacts described in the WHO global assessment of the burden of disease from climate change Presents overall results from this assessment, and describes their usefulness, and limitations, for informing policy 2

We Know that There are Many Important Links to Health Some expected impacts will be beneficial but most will be adverse Expectations are mainly for changes in frequency or severity of familiar health risks Health effects Temperature-related illness and death Extreme weather- related health effects Air pollution-related health effects Water and food-borne diseases Vector-borne and rodent-borne diseases Effects of food and water shortages Effects of population displacement CLIMATE CHANGE There is now little debate that climate change will affect health both directly and indirectly. Many studies and review papers have established the links between climate and health in qualitative terms. Based on Patz et al., 2000 3

But Policy-Makers also Want Quantification We want to know not only if health will be affected, but also How important are these effects? Which diseases could have the biggest impacts? Which populations are most at risk, and how? However, policy-makers also want clarification on how large a threat climate change poses to health, what are the largest disease risks, and who is most likely to be affected. It is therefore important to obtain the best possible quantitative assessment of the likely health impacts of climate change. This is a particularly challenging task. Compared to other environmental risk factors, climate range is a newly recognized phenomenon, with global scope, operating over long time periods and affecting an unusually wide range of health outcomes. The guidance presented here therefore outlines a general approach, and describes how the methods that were applied in the World Health Organization global comparative risk assessment project. It aims to help generate preliminary estimates of some of the health effects of climate change, and serve as a guide to developing more comprehensive and accurate assessments in the future. 4

Burden of Disease Assessment Burden of disease methods Use standardized approaches to provide quantitative mortality and morbidity information Use death and summary population health measures (e.g., Disability Adjusted Life Years — DALYs) Can be applied either to diseases (e.g., total burden from all sequelae of diarrhoea), or risk factors (e.g., the overall burden from all health effects of smoking, lung cancer, cardiovascular disease) in a defined population Can also inform on the distribution of burdens, by disease, population subgroup, etc. The disease burden of a population, and how that burden is distributed, are important pieces of information for prioritizing and defining strategies to protect population health. For policy-makers, disease burden estimates provide an indication of the current and future health gains that could be achieved by the targeted protection from specific diseases, and specific risks. To help provide a reliable source of information for policy-makers, WHO has developed methods to analyse the impacts of risks for health, and has estimated the impacts of 26 risk factors worldwide. 5

Estimates of Burden of Disease from Climate Change Completed at the global/regional level And at the regional/national level (Oceania) These approaches have so far been applied to climate change in two major studies. As assessment of the burden of disease from climate change at the global level (WHO, 2002; Campbell-Lendrum et al., 2003; McMichael et al., 2004), and at the regional/national level in Oceania (McMichael et al., 2003b). 6

Steps in Estimating Burden of Disease from Climate Change Step 1: Greenhouse gas emissions scenarios Time 2050 2100 Step 2: Global climate modeling: Generates series of maps of predicted future climate 2020s 2050s 2080s The general approach consists of the following steps: Selecting an appropriate set of scenarios of alternative possible futures (e.g., lower or higher rates of emissions of greenhouse gases, population growth), and the timescale over which to carry out the assessment Mapping the corresponding projected changes in climate properties

Steps in Estimating Burden of Disease from Climate Change (cont.) Step 3: Health impact model estimates the change in relative risk of specific diseases Step 4: Conversion to a single health measure 2020s 2050s 2080s Level Age group (years) 0-4 5-14 15-29 30-44 45-59 60-69 70+ 1 1.0 2 1.2 3 1.7 The general approach consists of the following steps: Modeling the impact on individual health risks (identifying the range of health outcomes that are both climate-sensitive and important in public health terms within the assessment population; quantifying the relationship between climate and each health outcome; linking the exposure measurement to the climate-health model). Using this information to calculate the climate change attributable burden of specific diseases. 2 1.2 3 1.7 1 1.0

Step 1: Defining Climate Scenarios Exposure scenarios used in the global assessment: Discrete climate scenarios derived from alternative future trajectories of GHG emissions 1961-1990 levels of GHGs and associated climate (baseline) Stabilization at 550 ppm CO2-equivalent in 2170 Stabilization at 750 ppm CO2-equivalent in 2210 Unmitigated current GHG emissions trends To calculate attributable and avoidable future burdens and the population at risk, the first step is to select plausible scenarios of the future, including changes in emissions of greenhouse gases, which are the main determinants of global climate change. Greenhouse emissions are distributed reasonably homogeneously throughout the atmosphere, and hence emission scenarios are usually defined at the global level. In the global assessment, a series of climate scenarios were used corresponding to different possible future paths of greenhouse gas emissions, leading either to continuous growth or stabilization of levels at different future time points. The IPCC (2000) has developed a series of 40 scenarios of plausible future trajectories for population growth, and economic and technological development [called the Standardized Reference Emission Scenario (SRES)]. These are used to estimate future greenhouse gas emission levels. Use of the SRES is currently recommended in national assessments to aid comparison between studies. 9

Projected Future Climate Change Business as Usual emissions 4 Temp increase (o C) 3 2 Baseline climate 1961-1990 1 The exposure is the output of global climate models that predict the effect of future emissions scenarios on climate properties, such as temperature or precipitation. Exposure is usually expressed as how much a climate property has changed from the agreed standard baseline condition (e.g. the average of the period 1961-1990, on the assumption that this period has not been strongly affected by human actions). A number of research groups produce global climate models that describe the projected changes in climatic conditions, and the geographical distribution associated with these different emissions. A series of six theme groups are approved by the IPCC. The IPCC website (http://www.ipcc.ch/) describes how to access the output from climate models. These slides show the output of the model used in the global assessment, for one important climate parameter – global average temperature – under the different scenarios of greenhouse gas emissions. 1900 2000 2100 2200 Year

Projected Climate Change with Emissions Stabilization 2210 Business as Usual emissions Stabilisation, 750 ppm 4 Temp increase (o C) 3 2 1 Stabilization scenarios describe alternative futures in which the level of greenhouse gases (usually described in terms of parts equivalent of carbon dioxide) are brought to a stable equilibrium level in the atmosphere, at a specific time point. These are then used as input to climate models that describe the climatic changes that are expected to accompany this trend in greenhouse gas levels. For example, this graph shows the expected path of global average temperature associated with greenhouse gases being gradually stabilized at a level equivalent to 750 parts per million CO2 (approximately 2.7 times pre-industrial levels), by the year 2210. Stabilization can be achieved by reduction in GHG emissions, protection and enhancement of “sinks” such as topical forests that take greenhouse gas emissions out of the atmosphere, and potentially (although controversially) through geo-engineering, such as air capture of carbon dioxide. 1900 2000 2100 2200 Year

Projected Climate Change with Emissions Stabilization (cont.) Business as Usual emissions Stabilisation, 750 ppm St. 550 ppm 4 Temp increase (o C) 3 2 1 Stabilization scenarios can be more or less ambitious. The green line here describes the expected temperature changes in a scenario in which GHG levels are stabilized at 550 ppm CO2 equivalent (approximately twice pre-industrial levels), by the year 2170. It is important to note that due to the long lifetime of GHGs, especially CO2, policies to stabilize emissions will only begin to have clear effects one or more decades from now. The aim of stabilization is therefore long-term, but nonetheless very important to avoid the increasing damages that are expected with higher degrees of climate change. For example, climate change is expected to cause some species extinctions even with relatively modest warming (e.g., less than 2°C from pre-industrial levels), but at values higher than 4°C, some 20-50% of species are expected to become extinct, and at above 5°C, a large fraction of ecosystems are expected not to be able to maintain their current form. 2170 1900 2000 2100 2200 Year

Step 2: Describing Climate Exposures Model outputs are usually supplied as gridded values of predicted changes in each climate parameter (temperature, rainfall, etc.) for each stabilisation scenario and future time period. For example, temperature in a single grid cell may be estimated to increase by 1.3ºC in 2030, relative to the baseline period. Some climate software (e.g., Schlesinger and Williams, 1997) provide estimates of climate changes at the national level. It is more common, and more accurate, to estimate climate changes at the level of individual grid cells, and use Geographic Information Systems (GIS) software to overlay these on digital maps of population distributions and administrative boundaries to estimate changes in exposure for specific populations.

Step 3: Selecting Likely Health Outcomes — Proposed Criteria Sensitive to climate Disease incidence should correlate with seasonal or intra-annual climate variation Important health impact Based on estimates of current mortality and/or morbidity Already modeled at an appropriate scale For example, existing models relating distribution of a disease to climate variables Health outcomes should be selected for which there is (1) evidence of sensitivity to short-term climate variability or geographic differences in climate, and (2) expected public health impact within the study population. Ideally, all health outcomes that are directly or indirectly linked to climate variability and climate change should be considered. In practice, the assessment is likely to be limited by (3) the availability of quantitative models describing climate-health relationships. 14

Examples of Important Climate-Sensitive Diseases Climate affects food production, water scarcity, and infectious disease transmission, which influence some of the biggest killers Already, each year: Undernutrition kills 3.5 million Diarrhoea kills 2.2 million Malaria kills almost 1 million These include diseases that have a direct physiological link with climate (e.g., cardiovascular disease), or infectious diseases where part of the transmission cycle for the pathogen occurs outside of the human host (e.g., vector-borne diseases and some diarrhoeal diseases). These diseases show a seasonal variation. Other impacts of climate change are more indirect, such as health threats arising via rising sea levels.

Availability of Studies that Estimate Effects of Climate Change on Health Health impact Available studies of climate change effects Thermal extremes Temperature-mortality relationships examined in multiple cities throughout world Extreme weather (floods, high winds, droughts) No complete analysis of linkage from climate change to changes in extreme events and health impact projections Diarrhoea Two local time series studies, no global model Malaria Three distinct global or continental models Dengue Two global models Malnutrition One global model of climate change to regional food availability A useful starting point is the list of outcomes considered in the global level study. These were deaths from cardiovascular disease in temperature extremes (incidence); diarrhoeal disease (incidence); malnutrition (prevalence); deaths in floods and landslides (incidence); vector-borne diseases (incidence of malaria); and people exposed to flooding from sea level rise.

Step 3: Modeling Climate-Health Relationships Health Impact 1: Diarrhoea The assessment process can be illustrated with an example from the global burden of disease study. Some level of diarrhoeal disease is present in all societies, so the issue is not whether climate change will cause the disease to spread, but whether it will increase in incidence in areas where it already occurs – particularly in developing countries where it is already a major burden. 17

Quantifying Climate-Diarrhoea Relationships Incidence of diarrhoeal disease is strongly related to climate variables. In Lima, Peru, diarrhoea increased 8% for every 1°C temperature increase. Checkley et al., 2000 Diarrhoea admissions Temperature This involves a statistical analysis of the effect of past variations in climate on disease either in space, or, as illustrated here, in time (e.g., measuring the effect of unusually hot or cold days on death rates). In such analyses, it is important to account for those non-climatic influences that would also affect disease rates, such as seasonal trends unrelated to climate or variations in socioeconomic conditions. Quantitative analysis can be used to yield an estimated change in disease rates, or the probability of disease occurrence, for each unit change in the climate variable (e.g., the increase in diarrhoea incidence per year for each degree Celsius increase in average ambient temperature). In this study, time series analysis suggests that diarrhoea increases 8% for every 1°C increase in temperature in this location. Other studies in developing countries generate similar results, suggesting that this may be a general pattern. Daily measurements Jan 1993 – Dec 1998

Converting to an Approximate Global Estimate Climate sensitivity 5% increase in diarrhoea per 1C temperature increase in developing countries Change in relative risk Projected temperature changes overlaid on population distribution map to give per capita increase in diarrhoea risk Disease burden attributable to climate change Relative risk under each scenario/time point multiplied by WHO estimates of current/future “baseline” diarrhoea burden in each region This can be used to calculate the relative risk (i.e., proportional change relative to the baseline) of each health outcome under each of the various future climate scenarios. Climate is geographically continuous, and values for the major meteorological elements provided are available as grid cells on a map. To measure the effect of these exposures on human health, it is usually necessary to have population data in a gridded format. Gridded maps of current population distribution, and projections out to 2015, can be obtained at a relatively high resolution, from 0.5 degree latitude and longitude grid cells (approximately 55 km2) down to 1 km x 1 km, e.g., from the Center for International Earth Science Information Network (CIESIN) Gridded Population of the World project. In general, population estimates decrease in reliability as the area of grid cells gets larger. Projections of future population size and demographic structure further into the future are usually available at the national level, either from national census agencies or from summaries developed by international agencies, such as the UN Population Division World Population Projects Database (http://esa.un.org/unpp/). Population projections will decrease in reliability as the projections extend further into the future. When appropriate maps of climate changes and population distributions have been obtained, GIS software can be used to overlay and link them, applying the appropriate relationships, as derived above. For example, combining the temperature change in each grid cell by the estimated sensitivity of diarrhoea to each degree centigrade increase in temperature, gives the estimated change in risk for that disease in that specific location. Applying this change in risk to the population of each grid cell, and averaging across the entire study population, will give an estimate of the average per capita change in diarrhoea rates within the population. 19

Converting to an Approximate Global Estimate (cont.) Climate sensitivity 5% increase in diarrhoea per 1C temperature increase in developing countries Change in relative risk Projected temperature changes overlaid on population distribution map to give per capita increase in diarrhoea risk Disease burden attributable to climate change Relative risk under each scenario/time point multiplied by WHO estimates of current/future “baseline” diarrhoea burden in each region The change in disease burden can be estimated by multiplying (1) the estimated relative climate change to the health outcome by (2) the total burden of disease that would have been expected to occur at the future time point, in the absence of climate change. The simplest assumption for the expected future burden of disease in the absence of climate change is that the disease burden will remain at current levels. In this case, the proportional changes in disease risk could be applied to current disease burdens, as measured by national statistics or from the WHO. It is more realistic, however, to take account of changes in other determinants of disease rates. Most fundamentally, these include future changes in population size and characteristics (e.g., age structure and degree of urbanization). It is therefore recommended to apply the estimated relative changes in risk defined above, to projections of future populations – while clearly differentiating between the effect of population changes and climate changes on disease burdens. Ideally, the estimates should also take into account, as far as is possible, the effects of other factors on underlying disease patterns. For example, malaria, diarrhoea, and other infectious diseases are expected to decrease with socioeconomic development, and technological improvements over time. This step was not carried out for the global assessment, as projections of expected future trends in burdens of specific diseases in the absence of climate change were not available at the time. WHO has, however, recently published such estimates at the regional level. The forward projections are based on general assumptions about the effect of projected changes in wealth, education, and application of new technologies, and the same trend is applied to all infectious diseases, with the exception of HIV/AIDS and tuberculosis. These are useful as a first approximation, and are currently being applied to future projections at the global level.

Converting to an Approximate Global Estimate (cont.) Health impacts attributed to the effect of climate change on diarrhoea in the year 2000: 47,000 deaths globally 23,000 in WHO Southeast Asia region World Health Organization, 2002 By applying the process above, i.e. Overlaying the projected changes in temperature from climate models on a population grid map to estimate exposure to increasing temperatures in each location Applying the temperature-diarrhoea relationships derived from the individual time series studies to each of the population grid cells (developing country populations only) to estimate per capita increases in diarrhoea due to temperature increases. Multiplying this by the WHO estimates of the burden of diarrhoea at the regional level in the absence of climate change. We estimate that the modest climate change that has occurred since the 1961-1990 period is already causing some 47,000 additional diarrhoea deaths per year in the year 2000 at the global scale, with approximately 23,000 occurring in the WHO South East Asian Region. 21

Step 3: Modeling Climate-Health Relationships Health Impact 2: Dengue Different modeling approaches are appropriate to different diseases. To take another example, there is concern over the extent to which climate change may increase not just the incidence of infectious diseases within populations, but their spread to other populations. This can be examined in the case of dengue. 22

Converting to an Approximate Global Estimate Climate sensitivity Relationship between climate variables and dengue distribution based on Hales et al. (2002) global model Change in relative risk Projected future climate scenarios applied to global model to map changes in disease distribution. Overlaid on population distribution map to give changes in population at risk (PAR). Disease burden attributable to climate change Percent changes in PAR applied to WHO estimates of “baseline” burden in each region (e.g., 50% increase in PAR assumed = 50% increase in mortality and morbidity). As for the diarrhoea example, this can be overlaid on maps of population distribution, in this case to estimate changes in the population exposed to the disease. The proportional changes can then be multiplied by the underlying burden of dengue (available, for example, from WHO) to estimate the changes in disease burden that may be attributed to climate change. 24

Converting to an Approximate Global Estimate (cont.) Dengue health impacts attributed to climate change in the year 2000: 1,000 deaths globally 25

Step 4: Aggregating Across Different Diseases All causes CVD* Floods Malaria Diarrhoea Malnutrition Subregion 19 1 5 8 AFR-D 36 18 9 AFR-E AMR-A 2 AMR-B AMR-D EMR-B 22 3 EMR-D EUR-A EUR-B EUR-C SEAR-B 80 7 52 SEAR-D WPR-A WPR-B 166 12 27 47 78 World One advantage of using standardized burden of disease approaches is that they allow for comparison and summation across different regions, and different diseases. In the global study, estimates were generated for a range of disease, across the 14 WHO epidemiological sub-regions. CVD* = Net changes in Cardiovascular disease deaths associated with both hot and cold temperatures

Conclusions: Poorest Populations are Most Vulnerable This allowed some broad conclusions at the global level. Climate change that has occurred since the period 1961-1990 may already be having significant health impacts – causing over 150,000 deaths, or the loss of over 5.5 million disability adjusted life years, annually, by the year 2000. This is the estimate just for the single year 2000, and the number is expected to increase over time as global warming continues and as populations increase. Some previously neglected causes may be more important than previously thought. Climate-change effects on malnutrition, diarrhoea, and vector-borne diseases appeared considerably more important than effects on flooding, or on deaths attributable to thermal extremes. Poor regions are probably impacted much more severely. Estimated DALY burdens per capita are several hundred times greater in the poorer regions of Africa, parts of the Eastern Mediterranean region, and Southeast Asia than in Western Europe, North America, and the more developed regions of the Western Pacific. Poor populations are at most risk because they already suffer a higher burden of the most important climate-sensitive diseases (malaria, diarrhoea, and malnutrition) these major climate-sensitive diseases mainly affect younger age groups. Health burdens from climate change appear to be borne mainly by children in developing countries.

But are Least Responsible for Causing Climate Change Cumulative emissions of greenhouse gases, to 2002 WHO estimates of per capita mortality from climate change, 2000 The study also allows us to show the need for a shared global response. This slide illustrates this using cartograms, which resize the country areas from their original dimensions according to the country-specific values of the variable being mapped based on its share of the world total of the variable. No change in country size vs land mass area would mean that the country has a share of the world variable equivalent to its proportion of the world's land mass. Scaling countries and regions according to their greenhouse gas emissions on one hand, and the estimated health impacts on the other hand, shows that it is those that are least responsible for climate change that are most vulnerable to its effects. The top graph expands or contracts the land area of countries in proportion to the amount of greenhouse gases that they have emitted up to the year 2000. (i.e. a measure of the effect of different populations on the global climate so far). The bottom graph expands or contracts the land area of countries in proportion to the estimated per capita mortality attributable to climate change in the year 2000, using the approach outlined earlier in the presentation (i.e. an estimate of health vulnerability to climate change). Map projections from Patz et al., 2007

“Adaptive” effects over time Limitations (1): Crude Representation of Non-climate Effects Health impact “Adaptive” effects over time Direct physiological effects of heat and cold Temperature associated with lowest mortality changes as temperature increases Diarrhoea Diarrhoea baseline drops, becomes insensitive to temperature if GDP per capita > US$6,000/year Malnutrition Future increases in crop yields from technological advances, trade liberalization, increased GDP Malaria Malaria baseline drops, developed regions remain protected from malaria invasion Disasters: Deaths in coastal floods Baseline and relative risk of deaths in floods drops with GDP Disasters: Inland floods and landslides Burden of disease assessments for climate change are useful in broad terms, but have some important limitations. These include the inevitable uncertainty around climate change and its effects on health, and particularly the effect of non-climate variables, which may modify (positively or negatively) the effect of climate change on health. This slide shows the assumptions that were made concerning the effects of adaptation to climate change over time, in the global study. These are more realistic than assuming that only climate will change in the future, but are necessarily very approximate.

Limitations (2): Many Impacts Cannot be Reasonably Modeled Examples Leishmaniasis, cholera, sleeping sickness, filariasis… Flooding impacts on diarrhoea, mental health, non-communicable diseases (NCDs)… Increased frequency of severe tropical storms Floods from melting glaciers, water shortages from melting glaciers Salination of water sources from sea-level rise Aeroallergens The most important limitation is that it is only possible to model some of the climate-health pathways quantitatively at this time. Estimates of burden of disease from climate change will therefore be incomplete. In presenting these findings to decision-makers, it is therefore important to make clear the limitations of these assessments: quantitative estimates are unavoidably uncertain; changes in non-climatic factors will influence both the baseline rates of disease and their sensitivity to climate effects; and many of the mechanisms by which climate change may affect health are not currently modeled, more likely leading to an underestimation rather than an overestimation of health threats. 30

Limitations (2): Many Impacts Cannot be Reasonably Modeled (cont.) Examples Forest fires Dust storms Effects on crop pests Effects via species extinction and biodiversity loss Social effects of population displacements The most important limitation is that it is only possible to model some of the climate-health pathways quantitatively at this time. Estimates of burden of disease from climate change will therefore be incomplete. In presenting these findings to decision-makers, it is therefore important to make clear the limitations of these assessments: quantitative estimates are unavoidably uncertain; changes in non-climatic factors will influence both the baseline rates of disease and their sensitivity to climate effects; and many of the mechanisms by which climate change may affect health are not currently modeled, more likely leading to an underestimation rather than an overestimation of health threats. 31

Interpreting and Using the Results Estimates of burden of disease from climate change are just one dimension of a health assessment. They should be presented alongside: Estimates of the underlying disease burden, irrespective of climate change How bad is the problem already? Information on the distribution of risks within a population Who is most vulnerable? It is therefore important to recognize that burden of disease assessments are not the only valid or useful approach to describing such impacts. Other kinds of information are also highly relevant, as listed here. 32

Interpreting and Using the Results (cont.) Qualitative assessment of potential health impacts that are difficult to quantify, or the possibility of extreme events What else could happen? Assessments of interventions What we can do about it? It is therefore important to recognize that burden of disease assessments are not the only valid or useful approach to describing such impacts. Other kinds of information are also highly relevant, as listed here. 33

Conclusions Global burden of disease study showed Much of the global burden of disease, especially in the poorest countries, is climate-sensitive. Failure to stabilize climate may already cause the loss of over 5.5 million years of healthy life (or over 150,000 lives) per year. This is expected to rise in future decades. There is a need for strengthened control of climate-sensitive diseases (in the short-term), and reducing climate change (in the long-term). The quantitative environmental burden of disease approach outlined here has a series of useful characteristics. First, by aiming at a comprehensive assessment, it gives a better representation of the health consequences of climate change than studies of single disease outcomes in restricted populations. Secondly, the quantitative approach helps to identify the relative public health burden of different climate-sensitive diseases in different populations. The global assessment, for example, showed that relatively small proportional increases in risk for climate sensitive diseases such as diarrhoea and malnutrition may cause very large increases in the total future disease burden. It also helped to demonstrate that the health risks of climate change fall mainly to children in developing countries, who have contributed least to the emissions of greenhouse gases that cause climate change. It therefore emphasizes the need for shared international responsibility for protecting health under a changing climate. 34

Conclusions (cont.) National burden of disease studies would Give more local, accurate, and context-specific measurements, with a stronger link to control interventions Provide a stronger basis for accessing global “adaptation funds” for health protection Such studies also clarify many of the limitations on current knowledge, such as the lack of quantitative information on many climate-health mechanisms. There is a need for wider application of these methods, within more national vulnerability and adaptation assessments, and more closely linked to adaptation policies. This should help strengthen the evidence base on what is necessary to protect health from climate change, and to access and apply the resources necessary to do so. Recommended reading: “ Climate change: Quantifying the health impact at national and local levels”, WHO, 2007 http://whqlibdoc.who.int/publications/2007/9789241595674_eng.pdf 35