PREDICTIVE RISK MAPPING OF WEST NILE VIRUS (WNV) INFECTION
The introduction of West Nile virus (WNV) into North America in 1999 sparked interest in predicting where and when the virus would appear next. New infections appeared to be geographically random, making it impossible to predict the location and timing of individual cases. It is possible, however, to identify areas of higher risk using geographical information systems (GIS), remotely sensed data (satellite imagery), ecological variables, and other spatial-analysis techniques. This approach has been useful in predicting the occurrence of other vector-borne diseases, such as Lyme disease and malaria.
Vector-borne diseases are particularly amenable to spatial and temporal analysis because they are highly influenced by annual seasonal variations in climate as well as unpredictable changes in climate and in the environment. Environmental conditions play a key role in determining the timing and intensity of the WNV cycle. Mosquito populations are especially sensitive to regular seasonal changes in climate and the environment, such as vegetation cover, rainfall, humidity, and temperature. The extrinsic incubation period, that is the time required from an infectious blood meal until transmission of the virus, is governed by temperature. Congregation of mosquitoes and bir. Environmental conditions affect the behavior of humans, which is particularly relevant when they spend time outdoors at peak periods of mosquito activity, such as dusk or dawn. These same conditions likely alter the behavior of horses and the humans who manage them. Although the source of WNV introduced in 1999 is not known, favorable conditions existed that allowed it to become established in the local mosquito and bird populations.
Defining the risk of acquiring infection with WNV is a key component of public health intervention strategies. Assessing WNV risk and deciding when and where to implement intervention strategies are largely based on surveillance of vector populations, suitable environmental conditions, and the presence of clinical disease in various host species. From the perspective of horse owners, knowing the risk of WNV in their area is a contributing factor to deciding whether or not to vaccinate. Surveillance of vectors and hosts over large geographic areas, however, can only produce maps with discontinuous patterns. Intervention decisions based on these maps alone require assumptions to be made about risk in the unsampled areas.
Predictive risk mapping is a 2-stage process in which components of the disease cycle (epidemiological, environmental, and/or entomological) are first used to build models. In the second step, predictions are made about the probability of membership within defined risk categories for all geographic sub-regions of the study
area, including those without existing surveillance data. Methods have become more practical and are applied to a broader range of diseases and study locations because remote sensing can now provide environmental information at the necessary spatial and temporal resolution. It is critical, however, to use appropriate geographic resolution, variables of interest, and methodologies cautiously for sound investigation of the stated hypotheses.
In 2002, West Nile virus (WNV) was identified in birds and horses in Saskatchewan (SK). A series of 4 observational studies were undertaken in 2003 to monitor the progress of this emerging public health issue. This paper describes how we used these data to then define and model areas of low, medium, and high risk of WNV infection for horses in SK. We then assessed the model’s predictive ability using 2005 data.