Figure 1.
Example of a simulated dataset containing an artificially created infectious disease cluster.
Example of a simulated dataset with an artificial cluster area (box) containing Nt treatment cases (black dots) and surrounding Nc control cases (grey dots) at the address level. Black squares with inscribed dots indicate ZIP code centroids to which address level cases would be spatially aggregated. Note: Nc cases drawn from underlying population of cases to retain actual spatial location and event date is randomized. Nt cases are randomly located within the cluster box area and event date is forced into a 3-month time period. Number of Nt cases is calculated as a function of the underlying at-risk population within the cluster box area and randomly chosen relative risk value.
Figure 2.
Effect of spatial resolution on power to detect significant space-time clusters.
Effect of spatial resolution on power to detect significant space-time clusters at significance level α = 0.05 for varying sizes of cluster radii of 0–16 km (A), 17–32 km (B) and 33–48 km (C). Each line represents disease case data aggregated to different spatial resolutions – the address level (solid line with solid squares), the ZIP code level (dashed line with open squares) and county level (dotted line with triangles). Relative risk (abscissa x-axis) describes the intensity of the artificially created clusters, where RR = 1 indicates the risk of a disease case occurring inside the cluster area is equivalent to that of occurring outside the cluster area (see Fig. 1). RR = 10 indicates risk is 10 times higher inside the cluster area relative to outside the area.
Figure 3.
Effect of spatial resolution on power to detect significant space-time clusters accounting for spatial precision.
Effect of spatial resolution on power to detect significant space-time clusters accounting for spatial precision for varying sizes of cluster radii of 0–16 km (A), 17–32 km (B) and 33–48 km (C). Each line represents disease case data aggregated to different spatial resolutions – the address level (solid line with solid squares), the ZIP code level (dashed line with open squares) and county level (dotted line with triangles). Relative risk (abscissa x-axis) describes the intensity of the artificially created clusters, where RR = 1 indicates the risk of a disease case occurring inside the cluster area is equivalent to that of occurring outside the cluster area (see Fig. 1). RR = 10 indicates risk is 10 times higher inside the cluster area relative to outside the area.
Figure 4.
Effect of spatial resolution on sensitivity.
Effect of spatial resolution on sensitivity, defined as the proportion of artificial observations included within the detected significant clusters for varying sizes of cluster radii of 0–16 km (A), 17–32 km (B) and 33–48 km (C). Each line represents disease case data aggregated to different spatial resolutions – the address level (solid line with solid squares), the ZIP code level (dashed line with open squares) and county level (dotted line with triangles). Relative risk (abscissa x-axis) describes the intensity of the artificially created clusters, where RR = 1 indicates the risk of a disease case occurring inside the cluster area is equivalent to that of occurring outside the cluster area (see Fig. 1). RR = 10 indicates risk is 10 times higher inside the cluster area relative to outside the area.
Figure 5.
Effect of spatial resolution on positive predictive value (PPV).
Effect of spatial resolution on positive predictive value (PPV), defined as the proportion of observations in the detected significant clusters and of the artificial cluster for varying sizes of cluster radii of 0–16 km (A), 17–32 km (B) and 33–48 km (C). Each line represents disease case data aggregated to different spatial resolutions – the address level (solid line with solid squares), the ZIP code level (dashed line with open squares) and county level (dotted line with triangles). Relative risk (abscissa x-axis) describes the intensity of the artificially created clusters, where RR = 1 indicates the risk of a disease case occurring inside the cluster area is equivalent to that of occurring outside the cluster area (see Fig. 1). RR = 10 indicates risk is 10 times higher inside the cluster area relative to outside the area.