This paper develops a large-scale Bayesian spatio-temporal binomial regression model to investigate regional trends in antibody prevalence to Borrelia burgdorferi, the causative agent of Lyme disease. Our model uses Guassian predictive processes to estimate the spatially varying trends and a conditional autoregressive scheme to account for spatio-temporal dependence. A novel framework, easily scalable to large spatio-temporal data, is developed. The proposed model is used to analyze about 16 million B. burgdorferi antibody Lyme tests performed on canine samples in the conterminous United States over the sixty-month period from January 2012 to December 2016. This analysis identified areas of increasing canine Lyme disease risk; prevalence of infection is getting worse in endemic regions and increases are also seen in non-endemic regions. Because Lyme disease is zoonotic, affecting both humans and dogs, the analysis also serves to pinpoint areas areas of increasing human risk.
Stella C. Watson Self1, Christopher S. McMahan1, Derek A. Brown1, Robert B. Lund1, Jenna R. Gettings2, Michael J. Yabsley2,3
1 Department of Mathematical Sciences, Clemson University, Clemson, South Carolina
2 Southeastern Cooperative Wildlife Disease Study, Department of Population Health, University of Georgia, Athens, Georgia
3 Warnell School of Forestry and Natural Resources, University of Georgia, Athens, Georgia
Self SCW, McMahan CS, Brown DA, Lund RB, Gettings JR, Yabsley MJ. A large-scale spatio-temporal binomial regression model for estimating seroprevalence trends. Environmetrics. 2018;29:e2538.https://doi.org/10.1002/env.2538