You are here:
DATA-DRIVEN DISCOVERY OF TEMPORAL AND GEOSPATIAL PATTERNS OF DISEASE TRANSMISSION: WEST NILE VIRUS IN MARYLAND
Citation:
Jorgesen, J., B. D'Ambrosio, J OrmeZavaleta, H. Luh, AND P. A. Rossignol. DATA-DRIVEN DISCOVERY OF TEMPORAL AND GEOSPATIAL PATTERNS OF DISEASE TRANSMISSION: WEST NILE VIRUS IN MARYLAND. Presented at Natural Science and Public Health: Prescription for a Better Environment, Reston, VA, April 1-3, 2003.
Description:
The necessity of rapid response to a developing disease outbreak often precludes systematic investigation of the mechanisms and patterns (temporal and geospatial) of spread. In order to deploy the most rapid response possible, we must exploit existing data to its maximum extent. These data often are observational in nature, collected during independent survey efforts. We demonstrate the use of relational Bayesian modeling (RBM), a model discovery technique developed using machine learning technology, to construct quantitative, biologically-consistent models of West Nile Virus spread from sparse survey data. Relational Bayesian modeling is a method for building models using relational data. It encourages the modeler to interact with the data and multiple hypotheses concerning the incidence and spread of the disease as a way to fully explore the combined data residing in multiple data tables. The models constructed may be updated as new information becomes available in the form of additional data or expert knowledge contributed by domain experts. Data on West Nile Virus cases in mosquitoes, horses, humans, and birds in Maryland collected during 2001, along with information on tire clean-up sites and collection facilities in Maryland, were evaluated using this technique. Our results indicate that RBM shows promise as a tool for assessment of spatial and temporal links in the epidemiological evaluation of disease transmission during the early stages of an outbreak or during an ongoing outbreak of a relatively rare disease.