Vector-borne diseases (VBDs) are the most common types of emerging infectious diseases and constitute as major threats to ecosystems. Ecological changes are frequently linked to emerging infectious diseases as alterations to vector population dynamics can increase the geographical distribution and abundance of pathogens. Understanding the interplay between the environment, vectors, pathogens, and humans that expedite pathogen population growth or range expansions remains a challenge. The rapid increase in availability of ecological information has led to “big data” opportunities and challenges in disease surveillance.
The goal of this study is to build geospatial models to predict vector and pathogen risk by accounting for ecological factors, using multiple analytical frameworks. We have previously identified some of the environmental features that have influenced population dynamics on I. scapularis, a vector of multiple important human diseases, using conventional regression methods. In this study, we focus on both the black-legged tick (Ixodes scapularis) and the pathogen for Lyme disease (Borrelia burgdorferi), the most prevalent vector-borne disease in the USA.
We briefly discuss the use of powerful statistical methods to analyze “big data” to identify influential factors. Using a more in-depth climate and environmental dataset, we investigate entomological, environmental, and host demographic factors that determine tick-borne disease risk. We use multiple regression, a Bayesian algorithm, and machine learning analysis. The Bayesian method allows us to identify factors that influence the population dynamics of tick populations due to its high precision and statistical power. The Bayesian approach addresses many of the limitations of multiple regression methods to identify variables of both practical and statistical significance. Machine learning is more flexible than classical regression methods and can be used in predictions.
Results/Conclusions
Preliminary coarse-scale analyses suggest that Lyme disease incidence rate is linked to both age and gender. The machine learning method appears effective for estimating tick prevalence in addition to the identification of important environmental factors. We will present these results as well as a comparison of the predictive power of these models at the ESA meeting in August 2018. Our preliminary studies demonstrate that these models effectively forecast vector and pathogen distribution over time and space with greater accuracy than conventional methods. With predictions of an increase in 2-3°C in global temperature by 2100, there is uncertainty in how tick-borne pathogens will shift and my research has implications for the mitigation of disease outbreaks.