concern is that if data are released at a finer scale it may be possible to identify individual patients.

Restricted access to individual health data necessarily makes detailed analyses of spatial patterns of disease more challenging, but it is an almost unavoidable consequence of privacy concerns. This restriction can be addressed in different ways. Seiler et al. (1999) evaluated the opportunity for septic contamination of groundwater by pharmaceuticals by using the specific spatial locations of groundwater wells and linking specific health conditions to individuals through their prescriptions. These authors were able to maintain individuals’ privacy by simply refraining from publishing well location data. In other cases, the spatial location (i.e., home address) of an individual may be known by the treating physician or responsible health official, but they would be prohibited from releasing the information (although curiously enough, addresses and even dispatch type for public safety responses involving ambulance or law enforcement are often published in community newspapers). Undoubtedly, making spatial attributes of epidemiological data available for research at appropriate scales and with patient privacy safeguards will continue to pose a challenge. One solution may involve the definition of a new data block of sufficiently small geographic size to be able to associate disease with geological phenomena while providing a sufficiently large error ring around an individual’s residence.

Federal agencies are increasingly using GISs at the interface of the earth sciences and public health. Examples include the Agency for Toxic Substances and Disease Registry (ATSDR), which was an early adopter of GIS (Cromley, 2003), and the Environmental Protection Agency(EPA) with its Toxics Release Inventory² (TRI). Although the TRI is not linked to disease data, there is potential to link to cancer registry data, asthma incidence and prevalence data, and other disease data that are spatially distributed. GIS has also been widely used for describing the distribution of natural hazards, for infectious disease modeling and outbreak investigations, for the detection of communicable disease clusters, and—with the recent concern about biowarfare—in new syndromic surveillance systems.³ Standard datasets collected by the National Center for Health Sta-