Abstract

Ehrlichia chaffeensis, the causative agent of human monocytotropic ehrlichiosis (HME), is maintained in a zoonotic cycle involving white-tailed deer (WTD, Odocoileus virginianus) as a vertebrate reservoir and the lone star tick (Amblyomma americanum) as the principal biologic vector. Using data from a prototypic white-tailed deer Ehrlichia chaffeensis surveillance system, we modeled the probability of E. chaffeensis occurrence using geostatistic analyses (kriging) and logistic regression. The analyses included the E. chaffeensis serostatus of 563 counties from 18 south-central and southeastern states. Cross-validation showed that kriging accurately predicted counties with high HME risk (87%). Large clusters of negative counties were accurately identified, but negative counties surrounded by large numbers of positive counties tended to be misclassified as high risk. Logistic regression modeling of the entire region and three subregions detected climatic and land-cover variables significantly associated with E. chaffeensis occurrence. The accuracy of each subregion model (78–85%) was higher than the regional model (75%). Use of subregions also greatly increased the specificity from 39% for the regional model to 48–68% for the subregional models. The predicted E. chaffeensis distribution had good concordance with human case data. The integration of a WTD surveillance system with geostatistic and logistic regression analyses was useful in developing HME risk maps.