Abstract
Existing data from multiple sources (e.g., surveillance systems, health registries, governmental agencies) are the foundation for analysis and inference in many studies and programs. More often than not, these data have been collected on different geographical or spatial units, and each of these may be different from the ones of interest. Numerous statistical issues are associated with combining such disparate data. Florida's efforts to move toward implementation of The Centers for Disease Control and Prevention's (CDC's) Environmental Public Health Tracking (EPHT) Program aptly illustrate these issues, which are typical of almost any study designed to measure the association between environmental hazards and health outcomes. In this article, we consider the inferential issues that arise when a potential explanatory variable is measured on one set of spatial units, but then must be predicted on a different set of spatial units. We compare the results from two different approaches to linking health and environmental data on different spatial units, focusing on the need to account for spatial scale and the support of spatial data in the analysis. We compare methods for assessing uncertainty and the potential bias that arises from using predicted variables in spatial regression models. Our focus is on relatively simple methods and concepts that can be transferred to the states' departments of health, the organizations responsible for implementing EPHT.
Mathematics Subject Classification:
Acknowledgments
The authors deeply appreciate the in-depth and constructive reviews of three referees, which led to a substantially improved article.
This publication was supported by the Florida Department of Health, Division of Environmental Health and Grant/Cooperative Agreement Number 5 U38 EH000177-02 from the Centers for Disease Control and Prevention (CDC). The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention.