COS 12-3 - A probablistic model of spatio-temporal disease dynamics: Urban cholera in Dhaka

Monday, August 8, 2011: 2:10 PM
17B, Austin Convention Center
Robert C. Reiner Jr., Entomology, University of California, Davis, Davis, CA, Aaron A. King, Ecology and Evolutionary Biology, University of Michigan, Ann Arbor, MI, Michael Emch, University of North Carolina-Chapel Hill, USA, Mohammad Yunus, ICDDR,B: Centre for Health and Population Research, Dhaka, Bangladesh, A. S. G. Faruque, International Centre for Diarrhoeal Disease Research, Bangladesh (ICDDR,B) and Mercedes Pascual, Ecology and Evolutionary Biology, University of Michigan,Howard Hughes Medical Institute, Santa Fe Institute, Ann Arbor, MI

The population dynamics of endemic cholera in urban environments, in particular the inter-annual variation in the size of seasonal outbreaks, remains poorly understood. The influence of climate variability has been addressed temporally but not spatially, despite the considerable socio-economic and environmental heterogeneity of increasingly large cities. Dhaka, the capital of Bangladesh, is a megacity in which cholera is endemic. With a probabilistic model, we address for Dhaka the role of spatial structure in disease dynamics, in particular its interaction with climate forcing by the El Nino Southern Oscillation (ENSO) and flooding. We utilize a multidimensional inhomogeneous Markov chain (MDIMC) model which we developed; this is a novel approach to disease modeling allowing for a large amount of flexibility in how spatial and temporal covariates can affect cholera dynamics.


We find a significant interaction between the climatological drivers and spatiality: their influence on cholera dynamics is locally differentiated with a subset of districts acting as a susceptible core, and with neighboring effects acting to propagate disease risk in space. The same model we use to test these effects can also be used to accurately forecast disease levels up to 11 months into the future. Additionally, using our MDIMC model we can estimate the distributions of these forecasts. MDIMC models have the potential to be useful for a wide variety of applications, not only for infectious diseases, but for other systems where both temporal and spatial effects are present.

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