Reconciling the Entomological Hazard and Disease Risk in the Lyme Disease System

Abstract

Lyme disease (LD) is a commonly cited model for the link between habitat loss and/or fragmentation and disease emergence, based in part on studies showing that forest patch size is negatively related to LD entomological risk. An equivalent relationship has not, however, been shown between patch size and LD incidence (LDI). Because entomological risk is measured at the patch scale, while LDI is generally assessed in relation to aggregate landscape statistics such as forest cover, we posit that the contribution of individual patches to human LD risk has not yet been directly evaluated. We design a model that directly links theoretical entomological risk at the patch scale to larger-scale epidemiological data. We evaluate its predictions for relative LD risk in artificial landscapes with varying composition and configuration, and test its ability to predict countywide LDI in a 12-county region of New York. On simulated landscapes, we find that the model predicts a unimodal relationship between LD incidence and forest cover, mean patch size, and mean minimum distance (a measure of isolation), and a protective effect for percolation probability (a measure of connectivity). In New York, risk indices generated by this model are significantly related to countywide LDI. The results suggest that the lack of concordance between entomological risk and LDI may be partially resolved by this style of model.

Keywords: Borrelia burgdorferi; Ixodes scapularis; Lyme disease; biodiversity; dilution effect; entomological risk; habitat fragmentation; land use; landscape; tick.

Figure 1

Simulated landscapes. (a) Modified Random Clusters landscapes with varying habitat (forest) occupancy and percolation probability. Forest cells are yellow, non-forest are white; (b) Simulated landscape with overlaid quadrats.

Figure 2

Model schematic and data. (a) Cartoon of model. The square represents the peridomestic region being considered. The gray circle represents an intersecting forest patch. When calculating entomological risk, the entire gray region is considered, as the patch’s entomological risk is a product of total area rather than area falling within the peridomestic region. When calculating exposure risk, only the portion of perimeter falling within the peridomestic region is considered (illustrated by bolded portion of perimeter); (b) Study landscape, including perimeter of 12-county region, 1.6 km buffer (black outer border in image), and intersecting forest patches (gray raster cells) derived from the National Land Cover Database 2011.

Figure 3

Lyme disease (LD) risk of simulated landscapes. Predicted LD risk as a function of habitat occupancy (forest cover). Each curve is evaluated for a landscape with a different percolation probability p.

Figure 4

Predicted Lyme disease risk as a function of mean minimum distance between patch edges, (a) across all simulated landscapes and (b) stratified by habitat occupancy A. Each curve connects landscape scores evaluated at sequential values of percolation probability p (0.05–0.5).

Figure 5

Predicted Lyme disease risk as a function of mean patch area. (a) LDI as a function of patch area across all simulated landscapes; (b) LDI as a function of area stratified by percolation probability p, with axes rescaled to better show maxima. Each curve connects landscape scores evaluated at sequential values of habitat occupancy A (0.1–0.9).