Epidemic West Nile virus encephalomyelitis: a temperature-dependent, spatial model of disease dynamics

Since first being detected in New York in 1999, West Nile virus (WNV) has spread throughout the United States and more than 20,000 cases of equine WNV encephalomyelitis have been reported. A spatial model of disease occurrence was developed, using data from an outbreak of serologically confirmed disease in an unvaccinated population of horses at 108 locations in northern Indiana between 3 August and 17 October 2002. Daily maximum temperature data were recorded at meteorological stations surrounding the study area. The distribution of the total number of degree-days elapsing between July 4 and the date of diagnosis of each case was best described by a normal distribution (mean = 5243 °F, S.D. = 1047). The days on which the average risk was >25, >50 and >75% were predicted (versus observed) to occur on August 23 (August 9), August 31 (September 2) and September 9 (September 9). The epidemic was predicted to occur 3 days earlier, or 4 days later, than observed if temperatures in the study area were uniformly increased, or decreased, by 5 °F, respectively. Maps indicated that WNV encephalomyelitis risk always remained greater in the northwest quadrant of the study area. Since WNV might exist at a hypoendemic level of infection, and occasionally re-emerge as a cause of epidemics in equine populations, by identifying factors that contributed to this epidemic, the potential impact of future epidemics can be reduced. Such studies rely on a GIS framework, availability of meteorological and possibly remotely sensed data and information on host and landscape factors. An early-warning system for WNV transmission in equine populations could be developed.