Any spatially distributed data can be analyzed using spatial statistics, and “spatial epidemiology” has developed as a subfield of epidemiology. Spatial approaches to understanding disease are now feasible because “the availability of geographically indexed health and population data, and advances in computing, geographic information systems, and statistical methodology, have enabled the realistic investigation of spatial variation in disease risk, particularly at the small-area level. Spatial epidemiology is concerned both with describing and with understanding such variations” (Elliott et al., 2000, p. 3). These authors suggest that there are four types of largely statistical and mathematical studies that fall under the rubric of spatial epidemiology—disease mapping, spatial correlation studies, risk assessment relative to point and line sources, and disease cluster detection. In addition, causal modeling could be added to this list.
Research at the interface of public health and the earth sciences is only as good as the data used to integrate the two. A wide range of geographical and geological data types, particularly remotely sensed data from earth observation satellites (e.g., Guptill and Moore, 2005), are readily available. Such data are, by definition, spatially distributed, and these data are geocoded to enable spatial modeling, geospatial analysis, and the use of GIS. The same cannot be said of most readily available epidemiological data.
The use of spatial techniques, including GIS and spatial analysis, requires that health data be available with their spatial coordinates. Although health data could be geocoded using either Universal Transverse Mercator spatial coordinates or simply the patient’s address, this has implications for the maintenance of privacy of an individual’s health status. In addition, a disease with a long latency or highly specific spatial data invites spatial artifacts (e.g., associating a disease with a residential location would be misleading for a work-related illness). In the reasonable absence of such specific information, data could be made available at the census block group or census tract levels. In reality, most health data—if available at all by location—are usually released by zip code (e.g., CHARS data for Washington State, which include detailed diagnostic and procedural information for each patient discharged from a hospital in the state) or by county (e.g., HIV/AIDS data from the Centers for Disease Control and Prevention). The CDC is very concerned with confidentiality, and the