Figure 1.
SatScan West Nile Virus (WNV) hotspot analysis and case/control selection.
WNV spatial clusters (i.e. hotspots) were determined based on 19 acute WNV human cases in 2000–2004. Only Cluster 1 was statistically significant at p<0.05. Additional cases (n = 81) were selected inside each hotspot weighted for the number of acute WNV human cases as follows (see Table 2): Cluster 1 (n = 33), Cluster 2 (n = 17), Cluster 3 (n = 11), Cluster 4 (n = 10), and Cluster 5 (n = 10). Control household locations (n = 100) were randomly selected from outside of the WNV hotspots.
Table 1.
Sources of environmental and socioeconomic data and the derived independent variables used in this study.
Figure 2.
Procedure for logistic regression model construction and West Nile Virus (WNV) human risk map development.
IVs - independent variables.
Figure 3.
Geographic distribution of West Nile Virus (WNV) human risk predictors with the corresponding spatial scales in the final logistic regression model.
Census variables were acquired at the smallest geographic unit available, i.e. census block to a block group, to census tract. The original independent variable shapefiles were converted to rasters in ArcMap GRID 30×30 meter format. The resulting grid layers were processed using Neighborhood function in ArcMap Spatial Analyst extension to produce smoothed output layers in which the value of each grid cell was a function of the cells within specified neighborhoods, i.e. the corresponding spatial scales of 0.5, 1.0, or 2.0 km (indicated by a number after each independent variable on the map). Similarly, each cell in distance grid layers (indicated as Distance to on the map) was assigned a value representing Euclidean distance to the nearest source cell.
Table 2.
SatScan™ WNV human cluster analysis and case selection.
Table 3.
Final logistic regression model for WNV human risk.
Figure 4.
Suffolk County West Nile Virus (WNV) human risk map based on the final logistic regression model.
WNV risk probabilities range from p = 0 (lowest) to p = 1.0 (highest). High WNV risk areas are defined as WNV human risk probability p>0.5. The geographic locations of acute WNV human cases in 2000–2004 (training dataset to generate the risk map) and 2005–2010 (validation dataset to assess the map accuracy) are shown. “No data” areas were generated due to incomplete census variables coverage in federal holdings (Plum Island, Brookhaven National Lab, parts of Fire Island National Seashore), privately owned islands, and along some sinuous coastlines.