Predicting the Lyme disease vector (Ixodes scapularis) from small mammals and the weather: A Bayesian approach

Date:

John R. Foster1; Shannon L. Ladeau2; Michael C. Dietze1

  1. Department of Earth and Environment, Boston University
  2. The Cary Institute of Ecosystem Studies

Abstract:

Lyme disease is the most prevalent tick-borne disease in the United States. The etiological agent, Borrelia burgdorferi, is transmitted between several mammalian reservoirs by ticks of the genus Ixodes. The most epidemiologically important host is the white-footed mouse (Peromyscus leucopus) as it’s an important blood meal host for larval Ixodes ticks, and the predominant reservoir for Borrelia. Furthermore, the white-footed mouse plays a critical role in both the survival and infection of nymphal Ixodes. Previous studies have shown that mouse population dynamics can help predict nymph abundance up to two years in advance. Here, we use Bayesian statistics to specifically compare tick population predictions among a suite of mouse population models, including a mouse only model, the minimum number of mice alive, and models including meteorological covariates, such as precipitation and temperature. Our models show that the tick-mouse dynamic is better predicted when meteorological variables, such as temperature and precipitation, are included as covariates. This was to be expected, as Ixodes development and fecundity are partially driven by both temperature and precipitation. The mouse only model, where we predicted ticks solely on the estimated mouse density, was prone to high degrees of error. The minimum number alive model performed about the same as the mouse only model. These results show that ticks, and therefore disease risk, can be accurately predicted from mice as long as meteorological variables are included as covariates. However, if this method were to be employed at other sites, density estimates should be used, as the absolute change in mice will differ across sites. Furthermore, the iterative nature of Bayesian statistics is ideal for estimating population density through time, because the current estimate of the population can be used to predict the future state. This work represents the first step towards forecasting near-term tick-borne disease risk.