Human Health in the Face of Climate Change: Science, Medicine, and Adaptation

"It's very difficult to attribute individual events to climate change which have an effect on health, but in the broad spectrum of these changes, clearly there is an association," Dye said. Statistical analysis can assign probabilities for particular phenomena; for example, one analysis found the 2010 heat wave in Russia was more likely than not a result of climate change. Scientists estimate that extreme weather events happen with fivefold greater frequency than they would in a hypothetical world without climate change.

Like the environmental effects, the health effects of climate change will not be evenly distributed. According to current projections, climate change will result in about 250 000 excess deaths worldwide by 2030, representing only 4% of the total deaths that year. Besides being relatively few and uncertain, the excess deaths will happen in the future. Policy makers tend to discount the importance of events that will occur long after their administrations have ended and may not be swayed by such predictions.

Measures that would prevent future climate-related deaths may have other benefits, however. Dye pointed to an intervention in which poor people without access to sanitary toilets were given subsidies to build latrines. The result was a drastic improvement in sanitation in the subsidized areas. "Not only [do] the people who get the subsidies [build toilets], but their neighbors and others in the communities take it up as well," he said. Improved sanitation should reduce the transmission of diarrheal diseases immediately, and make the communities more resilient to more frequent and severe flooding as the climate changes. "This is an example of being ready to ... address a health problem, both directly as it exists now and allowing for the future possibilities with respect to climate change," Dye said. Similarly, measures to improve health care and reduce pollution from fossil fuels can bring substantial co-benefits while also mitigating climate change or reducing its effects.

Dye closed with an overview of the ongoing efforts at the World Health Organization to keep health on the global agenda.

Mind the gaps: statistical issues in understanding climate and health

Peter J. Diggle of Lancaster University opened session one, on climate extremes and data, with a brief explanation of statistical model building. Using models, he explained, is like "buying information with assumptions." He cautioned researchers to incorporate statistical modeling into research at the experimental-design stage, rather than building a model after collecting the data. In general, scientists construct models on the basis of known phenomena, identify the variables most likely to influence future events, and then perform simulations to predict outcomes. Identifying the crucial variables at the start of a study will help investigators determine what types of data to collect to run the model.

That strategy creates some challenges for climate studies. "The relationship between climate and health has to be understood through synthesis of information from many different data sources ... often gathered on incommensurable scales," Diggle said. Nonetheless, he and his colleagues have successfully modeled climate effects on several diseases. In one study, for example, the team correlated rainfall and malaria rates, and identified the most effective anti-malaria interventions among several that public health officials had implemented. Another study, focused on leptospirosis, revealed that the locations of open sewers explained highly localized differences in infection rates within a city.

Sari Kovats of the London School of Hygiene and Tropical Medicine explained that incomplete data hamper statistical modeling, even in developed countries. In the UK, one of the first nations to mandate formal climate risk assessments at the national level, she has used government data to model the effects of extreme weather on health outcomes.

During the Europe's record-setting heat wave in 2003, mortality rates in London varied enormously between neighborhoods. It is hard to know exactly why. "We need to know about the housing characteristics [and] we need to know behavior within the household," Kovats said, adding that modeling is "limited by whether we can link our routine health data to these local or individual characteristics. That's something that's very, very hard to do."

There were lower mortality rates during floods than during heat waves, but Kovats cautioned that those data also had gaps. In particular, post-flood mortality rates may be lower because people move out of the flood-prone area and their subsequent illnesses and deaths are recorded in other districts.

Climate change makes extreme weather events more likely. (Image courtesy of Sari Kovats)

There are common themes between regions, however. As the climate moves toward greater extremes, the result will generally be an increase in health stresses (environmental conditions associated with poor outcomes), with the greatest effects on the most vulnerable populations. To anticipate those stresses, Luber and his colleagues at the CDC have developed a framework called BRACE (Building Resistance Against Climate Effects) to guide state and local adaptation efforts.

As Jane M. Olwoch of the University of Pretoria explained, local changes in climate do not have to be large to have enormous effects on health. Small shifts in rainfall and temperature can produce huge swings in the reproductive rates of mosquitoes and other important disease vectors. Declining biodiversity and changing migration routes among reservoir species can also push diseases into new areas. "This has far-reaching implications on the distribution and transportation of vectors and their diseases," she said.

Olwoch's team analyzed malaria rates and climate change in three provinces of South Africa. The results showed a correlation between temperature and malaria incidence, in contrast to earlier work in Botswana that identified rainfall as the main correlate of malaria. "But these are just simple correlations, which I think should undergo more serious, detailed analysis for us to be able to make conclusions," Olwoch said. She plans to look for data sources that provide more detailed information and to develop models to forecast future fluctuations in vector-borne disease.

Your local flucast

Jeffrey Shaman of Columbia University presented a modeling strategy for a disease linked more to seasons than to climate: influenza. Flu follows a predictable seasonal pattern in high latitudes, peaking in winter and declining in summer. Researchers have proposed several hypotheses to explain this pattern, but evidence for any of them has been hard to find.

After reanalyzing data from previous papers, Shaman's team realized that absolute humidity has a substantial effect on the survival of influenza virus in the air. With a mathematical model based on this phenomenon, the researchers showed that absolute humidity records track well with historical flu mortality rates across several U.S. states. They used the model in conjunction with data-assimilation methods and real-time observations to produce a forecasting tool for the flu season. "We're able to make predictions of what's going on with skill up to 9 weeks into the future," Shaman said.

Speakers from the first two sessions engaged in a lengthy interactive panel discussion with the audience. Topics included the co-benefits of climate change adaptations, which could make policy making more politically palatable, and the challenges of funding highly interdisciplinary climate and health studies.

Low absolute humidity favors influenza transmission. (Image courtesy of Jeffrey Shaman)

Making predictions

Andrew Dobson of Princeton University opened session three, on biodiversity and community health, with the observation that climate change is literally polarizing: Arctic and Antarctic conditions are changing much faster than conditions at lower latitudes. Besides making the poles good places to preview future climate disruptions, these changes add a sense of urgency to climate studies. "We've probably got about 25 years to study the Arctic before it's gone, [and] that's probably optimistic," Dobson said.

By tracking and modeling a nematode parasite of caribou, Dobson and his colleagues have developed a detailed understanding of how the Arctic's drastic climate change affects disease transmission. The researchers are now applying the model to other diseases of Arctic wildlife.

Meanwhile, new pathogens are arriving in the area. Dobson described the rapid spread of a new bacterial infection in musk oxen, resulting in massive die-offs of this already imperiled species.

Richard E. Paul of Institut Pasteur returned to the theme that small changes make big differences. While studying the biology of tick-borne Lyme disease, he built a model that could predict the incidence of the disease 4 months in advance in most years. But it failed spectacularly for two of the years he examined.

"When there was an anomalous warm winter and an anomalous cold summer, it was only 3 degrees' [difference], and the whole relationship of predictability was completely lost," Paul said. The milder winters and summers expected in some areas as the climate changes could shift the dynamics of Lyme disease dramatically. In another analysis, his team found dramatic variation in the distribution of cases of dengue fever across very small distances, and traced the effect to differences in the feeding habits of the mosquito vectors.

Robert P. Anderson of the City College of New York explained the nuances of building disease models that incorporate climate variables, and why the task a particularly difficult for vector-borne diseases. Models are mathematical constructs that allow researchers to make educated predictions; but models only work if the underlying data are sound and the model accurately represents the system's biology. Anderson outlined the principles his team developed for modeling ecological niches of disease vectors on the basis of physical and biological parameters. He illustrated the method using a model of Lyme disease, which includes data on ticks, populations of rodent reservoirs, and environmental factors that allow the disease to exist in an area. The team has produced software tools for researchers to use in other disease models.

The study tracks variables such as social cohesion, neighborhood walkability, cigarette smoking, and environmental pollution. Larkin's team examines air pollution, using environmental data to estimate emissions of different pollutants and then looking for correlations with respiratory and cardiovascular disease rates. Many of the pollutants are also greenhouse gases, highlighting the added benefits of measures to mitigate climate change.

After identifying health outcomes associated with pollutants, the researchers will be able to project the effect of climate change on each outcome. Larkin closed by inviting attendees with expertise in climate science or sociology to contact his group to explore collaboration.

Rachel Lowe of IC3, Spain, described her work on dengue fever in Brazil, which has the highest reported case counts of the disease. Epidemics of dengue fever generally occur between January and May, during the country's wet season. Lowe is developing a predictive model to provide health authorities with an early warning of impending dengue fever epidemics.

The model includes data on climate, demographics, socioeconomic factors, and dengue fever incidence for over 500 micro-regions of Brazil. After building the model to explain past epidemics, her team used it to develop thresholds for making predictions. "The problem that we often have when we're trying to model diseases is a lack of data. We can overcome this problem by introducing extra layers of uncertainty into the model using [a] hierarchical approach," she said. In a hierarchical model, unknown confounding effects, as well as measured covariates such as environmental and socioeconomic factors, can be accounted for.

Lowe and her colleagues tested the system before the World Cup soccer tournament held in Brazil in 2014, and she hopes to continue testing and developing the model to make regular dengue season forecasts.

Protecting community health

Marco Springmann of the University of Oxford turned the focus to nutrition-related health problems and the particularly devastating effects climate change may have on food systems. "Climate change has been called one of the biggest global health threats of the 21st century, and the health impacts related to food security may be indeed [among] the most important," he said. There are "[a] large number of people that might be affected."

The most obvious connection between climate and food is the potential for extreme weather events to cause crop failures and lead to undernutrition. However, poor diet composition causes 7 times more deaths than undernutrition. Unbalanced diet and obesity are now among the leading risk factors for death worldwide.

Projections of its effects on food production suggest climate change could lead to changes in dietary composition and weight status. Fruit and vegetable consumption, for example, may decrease. By 2050 these changes could lead to about 500 000 climate-related deaths relative to a reference scenario without climate change. Springmann argued that interventions to improve dietary quality would immediately improve public health, and also help countries offset the effects of long-term climate change.

Diet- and weight-related diseases could lead to about 500 000 climate-related deaths by 2050. (Image courtesy of Marco Springmann)

Joan Ballester of IC3, Spain, is studying whether a warming climate will decrease the number of deaths caused by cold weather in Europe. The answer depends on where one asks the question.

With data from 200 regions in Europe collected from 1998 to 2005, Ballester calculated the correlation between daily winter temperature and daily winter mortality for each region, finding that the tightness of the correlation varies dramatically from one region to another. Central and northern Europe show low correlations between death and cold temperatures, while the UK and southern Europe have much stronger correlations. "The interpretation is that those countries that have climatological winters that are harsher are those that are less susceptible to [cold] winter temperatures, essentially because they've taken steps toward acclimatization to these types of events," he said.

The data also revealed two types of acclimatization, suggesting that some regions are highly susceptible to daily but not to seasonal variations in temperature. In those areas, an anomalously cold day in a normally warm season could be deadlier than a series of cold days in winter.

A panel discussion featuring speakers from the third and fourth sessions ended the day, stimulating another round of questions and ideas from the audience. Asked about expanding her work, Lowe reported that she and her colleagues are building dengue fever models for different countries and studying malaria in Malawi. Other speakers fielded questions and comments about how to apply Arctic models to phenomena in the tropics, how heat waves may affect mortality, and how increasing fertilizer use may change projections of climate-related crop decline.

A project called the Great Green Wall of Africa takes the notion of combating climate change with vegetation further. The idea is to restore lost forests along a 15 km-wide belt across the continent, from Senegal to Djibouti. If successful, ecologists predict the reforestation could yield substantial benefits, such as reduced desertification and extensive agricultural and economic opportunities in multiple countries.

Simply salvaging degraded agricultural land is important. Restored farmland could feed more than a quarter of the 2.5 to 2.6 billion additional people the world is expected to have by 2050.

But nature-based solutions can incur costs; building parks in wet climates, for example, may expand mosquito habitats. Lindgren is optimistic that such costs can be mitigated; for example, by introducing biological controls such as natural mosquito predators into new parkland.

Lindgren closed with a plea for interdisciplinary collaboration to explore interactions between climate, health, and society. "If we understand these complexities, and by choosing the right solutions, we can actually not only avoid many of the local possible health effects of climate change, but we can also further decrease the local burden of disease," she said. "That will be very cost-effective for society."

The meat of the matter

Session five, on food and nutrition security, began with a presentation by Cassandra De Young of the UN Food and Agricultural Organization. She discussed the effects of climate change on oceans and fisheries, and the implications for nutrition. Seafood is among the most widely traded food products and is a particularly important source of animal protein in Africa and Asia. Fisheries often also serve as emergency food sources when crops fail. But data on fisheries are often scattered and inconsistent.

Changing ocean currents, rainfall patterns, and lake and ocean thermal structures, and ongoing ocean acidification, are causing major shifts in the global distribution of fish. In some areas, particularly in northern latitudes, fisheries may improve. However, some of the worst fishery declines are expected in many of the tropical regions most dependent on seafood. "Unfortunately the projections are fairly scary," De Young said, adding that in many areas it has proven difficult to implement scientists' recommendations for adapting to these changes.

Fish is a crucial source of high-quality protein. (Image courtesy of Cassandra De Young)

Matthew Baylis of the University of Liverpool began by polling the audience about their eating habits; the overwhelming majority of attendees relied on livestock for at least some of their food. Global data show that one third of human protein consumption comes from livestock and that demand is increasing.

Baylis and his colleagues are building mathematical models to understand how climate change will affect livestock diseases, over half of which are sensitive to climate. In different vector-borne diseases of cattle and sheep, the models show how various control measures would affect the progress of outbreaks. The models helped explain why in sheep bluetongue disease is relatively easy to control through a quarantine strategy, while Schmallenberg virus, carried to sheep by the same vector, spreads much faster. The models also demonstrate that vector-borne diseases can be exquisitely sensitive to small changes in climate.

Jose A. Centeno of the International Medical Geology Association discussed the biology and the toxicology of airborne dust. Most dust carried by winds comes from arid areas, particularly the Gobi and Sahara deserts. "This dust can travel the globe in a matter of days, perhaps weeks, and circle the entire globe," Centeno said.

Dust smaller than 2.5µm is especially important in health, because it can penetrate deep into the alveoli of the lungs. Though silica is the major component of airborne dust worldwide, depending on its origin dust can also carry toxic heavy metals such as arsenic, nickel, and cadmium, as well as pathogens.

Centeno described work linking earthquakes to outbreaks of valley fever, a dust-borne disease caused by Coccidioides immitis fungi endemic to the southwestern U.S. and parts of Mexico and Latin America. Meanwhile, U.S. soldiers returning from Iraq and Afghanistan have complained of a range of respiratory problems that may be linked to dust from that region.

Xavier Rodó of ICREA and IC3 turned the focus to Asia, where his group has studied a mysterious condition called Kawasaki disease. Epidemics of Kawasaki disease have been observed seasonally in Japan since the 1980s, but its causes are unknown. Climatological data revealed that outbreaks correlate closely with northwesterly winds blowing across Japan from mainland China.

Deeper analysis identified a possible source of the causative agent more precisely. "It was striking to see how all results converged to indicate that the potential origin ... came from an area in northeast China which is densely filled with crops," Rodó said. The researchers sampled the air over Japan on days with Kawasaki disease-associated winds, and performed a metagenomic analysis to identify potential pathogens. Numerous species of fungi appeared in the samples, but Rodó emphasized that the evidence linking any of them to the disease is still circumstantial.

Where does Japan's Kawasaki disease agent come from? (Image courtesy of Xavier Rodó)

In a panel discussion with the speakers from sessions five and six, Rodó expanded his discussion of Kawasaki disease, adding that since revealing the possible origin of the condition, he has been unable to get additional data from China. Other speakers discussed their own data collection problems, as well as the changing science of epidemiological modeling.

Accurate disease and climate surveillance

Madeleine C. Thomson of Columbia University began session seven, on climate-induced shifts in infectious diseases, by describing surveillance systems for infectious diseases and climate. An entomologist, Thomson became interested in weather stations while building malaria forecasting models that would account for climate change. She saw large gaps in the available climate data for Africa. Now, starting in Ethiopia, she and her colleagues are working to improve data availability. "The idea is to create high-resolution, high-quality data at the national level that is better than anything you can access elsewhere," she said.

The project deploys an integrated system that combines high-resolution ground station and satellite data, and then makes these data available to national meteorological offices. The researchers have helped build such systems for several countries.

The potential range of Marburg disease extends to Cameroon and Angola. (Image courtesy of A. Townsend Peterson)

A. Townsend Peterson of the University of Kansas talked about the questions underlying his efforts to map disease spread: "For a given disease system, where do we expect it to occur, and where do we expect it not to occur?" The answers are not simple and not always popular.

Peterson cited the example of Marburg virus, for which he and his colleagues built ecological and distribution models in early 2004. The models predicted that the virus could occur well outside its historical range in East Africa, even into northern Angola. "I was ... mocked at a meeting for that northern Angola prediction, because everybody knew that Marburg was an East African disease," Peterson said. But an outbreak of Marburg virus occurred in Angola in October the same year.

Peterson's team has since studied the ranges of several other diseases, showing that climate change is already affecting pathogen spread. Not all the changes in distribution will be negative for humans, however.

Mercedes Pascual of the University of Chicago reiterated earlier speakers' contention that while climate change is global, its effects on diseases will vary dramatically between regions and localities. Pascual described her work studying malaria in Ethiopia and Colombia and cholera in Bangladesh.

In Ethiopia and Colombia, Pascual's team found that malaria incidence clusters according to altitude. "As temperature decreases with altitude, we have a decrease in malaria and a change in the dynamics [of the disease]," she said. In warmer years, malaria cases increase at higher altitudes, mimicking what the investigators anticipate will happen with global warming.

Malaria decreases with altitude; cholera increases with urban poverty. In Bangladesh the team found drastically higher rates of cholera in the dense inner-city areas of Dhaka than in the city's periphery. "My prediction would be that we are going to go into worse scenarios in these very large cities that are growing at a very fast pace," she said.

Deirdre Hollingsworth of the University of Warwick ended the session with a discussion of neglected tropical diseases, which affect the "bottom billion" poorest people in the world. Populations at the highest risk for these diseases are often the same ones most vulnerable to rising sea levels and other effects of climate change.

Hollingsworth develops disease models to maximize the efficiency of treatment interventions for these neglected diseases. In one project, her team modeled the epidemiology of the intestinal worm Ascaris lumbricoides to determine the best strategy for distributing antihelminthic drugs. The model suggested that the timing of drug delivery could be critical to the effectiveness of a treatment campaign. But the researchers have limited data on the relationship between egg survival and climate, so she cautioned that using the model to suggest specific policy changes for timing drug delivery is "a long way from where we are at the moment." Although their roles in neglected tropical disease transmission are not fully understood, both weather and climate are likely to influence the long-term success of programs to control disease.