Detecting spatiotemporal associations between Lyme disease and environmental factors in New England

Keywords: Lyme disease, environmental factors, spatiotemporal heterogeneity, feature selection, spatiotemporal weighted regression (STWR)
Abstract Type: Paper Abstract

Abstract

Landscape fragmentation and climate factors may be related to the risk of Lyme disease (LD), as they may have an impact on the habitats of ticks and hosts, as well as interactions among ticks, hosts, and humans. Most previous studies on the environmental determinants of LD focused on particular land cover classes (e.g., forests) or climate factors (temperatures and precipitation) and were conducted at coarse spatiotemporal scales, typically at the county level over a specific period of time. Using data mining techniques, this study examined the spatiotemporal association between a variety of environmental factors and LD in the New England region of the United States during 1990-2019. First, we used an ensemble feature selection model to select the most important risk factors from thousands of metrics, and then we used spatiotemporal weighted regression (STWR) to detect spatiotemporal associations. The results of the study suggest spatiotemporal heterogeneity in the associations, which reveals both consistency and variability, as well as uncertainty in the associations. The findings could be useful for modeling and predicting disease risks.