COS 88-5 - Epidemic malaria dynamics and rainfall variability in North West-India

Thursday, August 6, 2009: 9:20 AM
Galisteo, Albuquerque Convention Center
Karina F. Laneri1, Mercedes Pascual2, Anindya Bhadra3, Edward L. Ionides3, Menno J. Bouma4, Ramesh Dhiman5 and Rajpal Yadav6, (1)Ecology and Evolutionary Biology, University of Michigan, Ann Arbor, MI, (2)Ecology and Evolutionary Biology, University of Michigan,Howard Hughes Medical Institute, Santa Fe Institute, Ann Arbor, MI, (3)Statistics, University of Michigan, Ann Arbor, MI, (4)Infectious and Tropical Diseases, London School of Hygiene and Tropical Medicine, University of London, London, United Kingdom, (5)National Institute of Malaria Research, Delhi, India, (6)National Institute of Malaria Research, Nadiad, Gujarat, India
Background/Question/Methods

Epidemic or “unstable” malaria occurs at the edge of the distribution of the disease in highland and desert fringes, where temperature or rainfall limit transmission. Thus climate variability has the potential to play an important role in driving malaria cycles in the size of epidemics at interannual time scales. There is indeed already some evidence for such a role in highlands of Western Kenya and Southwest Uganda. However, the alternative explanation of intrinsic cycles, driven entirely by patterns of immunity, was also proposed for epidemic malaria (and dengue).

Here, we address these alternative explanations with extensive temporal records of cases in desert fringes of North-West India, and a recently developed statistical inference method for stochastic differential equations. Malaria outbreaks in these regions can result in several thousand reported cases per month. We find a strong correlation between total monsoon rainfall and reported monthly cases from 1986 to 2006. We further consider a stochastic SEIRS (Susceptible-Exposed-Infected-Recovered-Susceptible) model that incorporates in a mechanistic way a simplified version of mosquito dynamics. The model is fitted to the time series data for three different districts using the sequential Monte Carlo method (MIF, Maximum Likelihood via Iterated Filtering) developed by Ionides et al. (2006).

Results/Conclusions
Results indicate that rainfall forcing can generate biennial periods, and can interact with disease dynamics to determine the timing of longer periods present in the data. Thus, rainfall plays a key role in the interannual variability of the disease and this role is robust to consideration of disease dynamics, including immunity.

Copyright © . All rights reserved.
Banner photo by Flickr user greg westfall.