Temporal patterns in ecological systems can be extremely dynamic, and as such, difficult to understand and predict. The abundance of species can fluctuate seasonally and can also differ in size from year to year, with cycles that repeat with some regularity or fail to do so and are irregular and largely unpredictable. The outbreaks of many infectious diseases can exhibit a rich array of complex temporal patterns. An understanding of these patterns is critical to evaluate control efforts. Prediction of these patterns can also help us to implement control and treatment in a timely fashion by anticipating large epidemics.
I am particularly interested in the temporal and spatial variability of infectious diseases, such as cholera and malaria, that are climate sensitive. Our work addresses the role of climate variability (e.g., rainfall, the El Niño Southern Oscillation [ENSO]) in the year-to-year variation in the size of epidemics. We seek to understand the role of climate in the context of the dynamics of the disease itself. As the result of the processes of transmission and the buildup of immunity in the population, disease dynamics can generate cycles in the size of epidemics over time, in the absence of any climate variability. These cycles can in turn interact with climate influences in ways that defy simple analyses and limit predictions based on climate alone.
Long temporal records, or time series, from surveillance programs and hospital records are invaluable to address these limitations and disentangle the respective roles of climate and immunity. Our work identifies regularities in the patterns of epidemics over time and relies on mathematical models of disease transmission to understand these past patterns. Perhaps more challenging, we can ask whether understanding the past allows us to forecast the short-term future and develop an early-warning system.
The comparison and fit of epidemiological models to time series data is not easy because we lack information on all the relevant variables that are part of the system. We are fortunate when we have a single long record on the number of weekly or monthly cases over several decades. For example, thanks to the invaluable efforts of the International Centre for Diarrhoeal Disease Research in Bangladesh (ICDDR,B), cholera records exist from 1966 to the present for a rural area south of Dhaka and from 1980 to the present for Dhaka itself. Other important variables are not, however, typically measured or can't be measured over time, such as the levels of population immunity or, for a vector-transmitted disease such as malaria, the abundance of the mosquito vector. Thus, the disease systems we wish to understand are a bit like a "black box" for which we have only partial information and mostly hidden variables. With our mathematical models, we seek to open this black box and make inferences from the data on the main mechanisms that are at play.
The oceans have a special place in our studies since we view them as an important link between disease and climate at a global scale. Phenomena such as ENSO in the Pacific can drive changes in regional weather throughout the globe. Our work considers global climate variability and looks for the existence of "hotspots" in oceanic variation (specifically in sea surface temperatures [SST]) associated with disease patterns. These can give us a handle on prediction because of the inherent lags between changes in SST far from the region of interest and disease responses to local variability in the weather.
Besides our work on cholera in Bangladesh, we are also addressing similar questions on the dynamics of epidemic malaria in African highlands and in desert and semidesert regions of India. These are known as regions of "unstable" malaria because at the edge of the spatial distribution of the disease, temperature or rainfall limits the abundance of mosquitoes or the development of the parasite. As a result, large epidemics occur intermittently, with high morbidity and mortality because of a lack of constant exposure to the disease in these populations.
These transition regions, especially highlands, bring us to another area of our research on the role of climate change—and not just climate variability—on disease patterns. Here our focus shifts to long trends, such as those in warming temperatures. Increasingly, the temporal and spatial patterns of infectious diseases must be understood in the context of long-term change, not just in the environment but also in socioeconomic conditions and in the evolution of the pathogens themselves. Our current work is addressing, for example, whether the exacerbation of malaria in recent decades in African highlands is already in part the result of warming temperatures. With our mathematical models, we would also like to understand the interplay between climate change and the evolution of drug resistance.
Finally, pathogen evolution and its interaction with epidemiological dynamics are important to another disease closer to home, influenza. With a combination of mathematical models and analyses of epidemiological time series and genetic sequence data, we hope to contribute to the understanding of how the pathogen's evasion of the immune system influences disease dynamics at the population level, and vice versa.
A general theme running in the background of all this is how much complexity must we incorporate in our mathematical models to address the complex temporal and spatial patterns of infectious diseases. There is always an unavoidable trade-off between complexity and tractability, between levels of detail and understanding or predictability. This makes for fascinating questions on the relevant levels of detail and on simplifying complex biological models while retaining information on those relevant details.
