Implications of Climate Change
on Malaria in Karnataka, India
Senior Honors Thesis in Environmental Science
Center for Environmental Studies, Brown University
by Colleen Reid
Readers: Harold Ward, Brown University
Anthony J. McMichael, London School of Hygiene and Tropical Medicine
Andrew Foster, Brown University
[ Summary ]
Malaria causes 300 million new infections and 1 million deaths a year (WHO, 1999). Malaria is the second most fatal communicable disease and is a public health problem in 90 countries in the world, where 40% of the human population live (WHO, 1998). Attempts in the 1950’s and 1960’s to eradicate malaria globally have been abandoned as malaria vectors have become more resistant to insecticides and the parasites that cause malaria are becoming resistant to chloroquine and possibly other anti-malarial drugs, making prevention and treatment increasingly more difficult and costly (Sharma, 1996a). Meanwhile, there is a growing consensus that anthropogenic actions will change the earth’s climate in the near future and that it may already be happening. The Intergovernmental Panel on Climate Change (IPCC, 1996) predicts a 1-3.5°C rise in global mean surface temperature by 2100, and many scientists predict that the effects on the hydrological cycle will create more extreme weather events such as hurricanes, floods and droughts (Patz et al., 1998). Human health is expected to be affected by climate change, both directly through increased mortality from extreme temperatures and weather events, and also indirectly, through effects on morbidity and mortality related to changes in food production, exacerbated air pollution, demographic displacement, and changes in the distribution of biological organisms that transmit vector-borne diseases (McMichael et al., 1996). These diseases, because of the dependence of the vectors and pathogens on climatic factors, are expected to change in distribution and intensity. Malaria is one of the vector-borne diseases that is expected to be most sensitive to long-term environmental change (WHO, 1999).
Every malarious area in the world has its own particular malaria ecology depending on its vectors, parasites, vegetation, host population and a variety of other factors. By analysis of the relationship between past climatic changes and malaria, it is possible to begin to anticipate what effect future climatic changes might have on malaria in that region. In this study, I assessed the level of correlation between mean temperature and precipitation, and malaria incidence rates in Karnataka, India from 1975-1984 and 1986-1995.
Temperature, precipitation, relative humidity, and wind are the four main climatic factors that affect malaria transmission and upon which the predictions of the effects of climate change on malaria are based. These relationships can be best understood in relation to the malaria life cycle. There are maximum, minimum and optimum temperatures for the development and survival of both the parasite and the vector of malaria, and increases in temperature tend to show increases in feeding and egg laying frequency of the vector. The amount of precipitation affects the amount of surface water within which the malaria vectors, different member of the Anopheles family, can breed. Relative humidity limits vector survival, and strong winds hinder biting by Anopheline mosquitoes and also allow for the distribution of the vector further than its own short flight span.
Changes in temperature, rainfall, and relative humidity due to anthropogenic climate change are expected to influence malaria directly by modifying the behavior and geographical distribution of malaria vectors and by changing the length of the life cycle of the parasite (Martens et al., 1995). Climate change is also expected to affect malaria indirectly by changing ecological relationships that are important to the organisms involved in malaria transmission (the vector, parasite, and host). Examples of such indirect forces are deforestation and habitat changes due to climate change that may affect which species of Anophelines are able to survive (McMichael and Githeko, 1999). Recent evidence shows that changes in temperature and precipitation have already changed the distribution and behavior of malaria. Many time-series studies and studies of epidemics have been done to find explanatory variables for changes in malaria transmission, but many of them do not take into account climatic factors. Many studies have reported the relationships between factors other than climate that affect malaria rates such as urbanization, migration, irrigation, agricultural practices, deforestation, and malaria control efforts.
Less research has been published on climate change and malaria in India than in other parts of the world, despite the fact that malaria is still prevalent, the population is very large and growing, and India has such a long history of malaria control efforts. Malaria is still a serious disease in India as 20,000 people and an estimated 577,000 DALYs (disability-adjusted life years) were lost due to malaria in 1998 (WHO, 1999). India has also spent a considerable amount of money on malaria control operations through its National Malaria Eradication Programme (NMEP). In recent years, increases in malaria incidence have been attributed to resistance of the mosquitoes to pesticides and of the parasites to anti-malarial drugs, thus limiting the effectiveness of malaria control attempts.
In this study, I have assessed the degree of correlation between two measures of malaria incidence, API (Annual Parasite Incidence) and SPR (Slide Positivity Rate), and mean temperature and precipitation for eighteen districts in the state of Karnataka, India. The malaria data were collected and published by the NMEP, and the climate data were taken from a climatology created by the University of East Anglia in Norwich, UK. I focused on the years from 1975 through 1995, excluding 1985 for which the NMEP has not published malaria data. There are sources of error in both the malaria data and the climate data, however, for the purposes, time-frame, and extent of this study, these data are sufficient for a preliminary analysis of the relationships between climate and malaria in Karnataka.
I chose to study Karnataka after doing an analysis of district level API (Annual Parasite Index) trends for fourteen states within India. When looking at the district level patterns within each state, I found that the state patterns fell into four categories. One pattern exhibited high API levels in the 1970’s that decreased significantly in the mid 1980’s and never increased. Another showed low API levels throughout, while others had API rates that followed India’s malaria control history. One district’s API levels that stayed the same throughout. Karnataka followed the malaria control history in India, which allowed for the study of the effects of climate on malaria while malaria control was successful, and also in the 1990’s when malaria increased despite control efforts. Karnataka is also interesting because of the heterogeneity of the API trends between districts and of the different climatic regions within the state due to the effect of topography.
My methods involved compiling the data by district, and then performing simple linear and multi-linear regressions on different sets of climatic variables and API and SPR. I used both API and SPR because they are both rates of malaria prevalence, but there is some uncertainty over which rate, API or SPR, is a better measure of malaria. I calculated the Pearson’s correlation coefficient between mean temperatures in cold months and hot months, and the Pearson’s correlation between mean temperature and precipitation within months to assess the extent to which mean temperature and precipitation within a month and between months are auto-correlated in Karnataka. I found the specific multi-variable linear regression that best predicted each malaria rate by district. I then performed regression analyses on groups of districts that had similar API trends, on the change in API and SPR rather than the actual values, and on data that excluded years of effective malaria control.
API and SPR show similar trends from 1975 - 1995: high malaria levels in the 1970’s decreasing to low levels in 1985 and then increasing in some districts again in the 1990’s. The Annual Blood Examination Rate (ABER) remained above 10% in most districts. The average yearly mean temperature increased in all districts of Karnataka from 1970 to 1998, whereas the precipitation trend was variable.
January mean temperature, April mean temperature, May mean temperature, December mean temperature lagged one year (December mean temperature correlated with malaria of the next calendar year), April precipitation, and November precipitation all significantly correlated with API and SPR in the most number of districts. The high number of correlations for these months is logical in terms of the climate of Karnataka. December and January are the two coldest months, and April and May are the two hottest months. Intuitively, variation in the mean temperature in these months is most likely to cause changes in malaria rates because these are the months of the hottest and coldest temperatures, which are the most likely to reach extremes that could limit vector and parasite survival and behavior. April and November are the two months at either end of the monsoon season and are thus the most likely to have enough variability in precipitation to change whether mosquitoes can breed.
The correlation coefficients for the linear regression between all the monthly mean temperatures and API or SPR were negative. This makes sense in the hot months, in which an increase in temperature would limit vector and parasite survival and therefore cause a decrease in malaria rates. In the cold months, however, this finding implies that an increase in mean temperature will bring a decrease in malaria incidence. The strength of the correlations is very high with many districts' correlations significant at the 99% level. This makes this finding all the more interesting, because logical reasoning and results from other studies would conclude that an increase in temperature in cold months should cause an increase rather than a decrease in malaria rates.
The most likely explanation for the finding that increases in mean temperatures in both hot and cold months is correlated with a decrease in malaria prevalence is the autocorrelation between mean temperature in hot months and cold months. This hypothesis is supported by the finding of positive Pearson’s correlation between the mean temperature in the cold and hot months. This implies that an increase in cold monthly mean temperatures can cause a decrease in malaria by way of an increase in hot monthly mean temperatures. However, this correlation could also imply that an increase in mean temperature in hot months may cause a decrease in malaria rates via a decrease in mean temperature in the cold months.
Another possible hypothesis to explain the negative correlation coefficient involves the relationship between temperature and relative humidity. For a given amount of moisture in the air, an increase in temperature causes a decrease in relative humidity (Webb, 2000, personal communication), which can limit Anopheles survival. I could not test the validity of this hypothesis for Karnataka because I did not have access to reliable district level relative humidity data.
The correlation between mean temperature and precipitation may also lead to an explanation of the negative correlation coefficient between mean temperature and malaria prevalence rates. Mean temperature and precipitation in January, April, and May, had negative Pearson’s correlation coefficients, whereas November and December had positive Pearson’s correlation coefficients. The negative correlation between mean temperature and precipitation in January implies that hotter Januarys tend to also be even drier than the average January, and thus, this increased dryness may be the limiting factor for malaria transmission. However, the negative correlation between December mean temperature and December precipitation does not follow this reasoning therefore limiting the viability of the idea that the relationship between temperature and precipitation can explain the negative correlation between mean temperature and malaria prevalence in cold months.
The coefficients for significant regressions between monthly precipitation and API or SPR are mostly positive, except for two districts, although the magnitude of these coefficients is rather low between ±1. The R2 values show that malaria rates and precipitation are correlated, but the correlation coefficients imply that a change in precipitation will likely be associated with a change in malaria prevalence.
The multi-linear regression results show that the amount of variation in SPR or API that can be predicted by a specific combination of climate variables, different for each district, ranges from 30% to 87%. There is also a geographical clustering with respect to the amount of variability predicted by climate variables alone, and with respect to the influence of mean temperature versus precipitation on malaria incidence rates. The geographic clustering is likely due to different climatic regions within the state or possibly to specific malaria ecology regions.
I performed regression analyses on groups of districts based on the API trend they exhibited to assess whether this would increase the predictive power of climate variables on malaria rates. However, this did not improve the explanatory ability of the regression model. Neither did regression analyses after removing data from years in which API was low, presumably due to years in which malaria control efforts were effective. Analysis of change in API or SPR and monthly mean temperature and monthly precipitation did not correlate as highly as did regular API and SPR rates.
This study found that malaria rates and climatic variables were significantly correlated in the districts of Karnataka, India. There was regional clumping with a greater percentage of the variability in malaria rates predicted by climatic factors in the center interior of the state, and lesser in the north, south, and west. Also, the southern interior districts’ malaria rates were more significantly correlated with precipitation than mean temperature, whereas the Northern districts’ malaria rates were more significantly correlated with mean temperature than precipitation. The geographic clustering allows for identification of areas that are more vulnerable to changes in malaria rates due to changes in climatic variables. However, the correlation coefficients implied that increases in mean monthly temperatures are associated with decreases in malaria incidence. Therefore, any change in mean temperature and/or precipitation is likely to have effects on malaria incidence in Karnataka. These results can have implications for climate change policy and for malaria control efforts. Further studies that account for all possible confounding factors and that are done at a smaller spatial scale, will improve our understanding of which areas will be most affected the most by changes in climate.