Superfund Research Program

New Methods of Spatial Analysis Are Developed for Environmental Epidemiological Data Using a Geographic Information System

Release Date: 10/29/1997

Epidemiological studies of hazardous waste problems routinely confront inherent difficulties. For instance, the populations that live near hazardous waste sites are usually relatively small; however, some of the health outcomes, such as birth defects and cancer, are relatively rare, and hence require information from large populations. Accurate and precise exposure information is also usually lacking. These kinds of information limitations often make it difficult to draw strong conclusions about increased disease incidence in populations that live in the vicinity of hazardous waste sites. New approaches that effectively overcome these limitations are needed to evaluate the data that is available in environmental epidemiological studies.

Advances in computer technologies provide sophisticated tools that can aid in the management and analysis of epidemiological data. Researchers at Boston University are currently exploring the means by which geographic information systems (GIS) can be used to develop new methods of spatial analysis for epidemiologic data. GIS facilitate the integration of diverse kinds of environmental and health information in the context of a map, making it possible to perform certain kinds of epidemiologic analyses that were previously unfeasible. An ambitious GIS database with unusually complete environmental information is being developed with existing and pending case-control data from the Upper Cape Cancer Incidence Study to address the issue of elevated cancer incidence in Cape Cod, Massachusetts.

Scientists have been digitizing pertinent spatial (residence and exposure zones) and aspatial (years of residence, age, and sex) data from the Upper Cape Cancer Incidence Study. In this ongoing effort, thousands of cancer case and control residences have been entered into the environmental GIS along with specific medical and personal information that was obtained in extensive field interviews. This information has been used to produce new kinds of "cancer maps." The purpose of these maps is to reveal any cancer hotspots and relate them, if possible, to environmental features of the area. Two new techniques have been developed for this purpose: odds-ratio maps and adaptive rate smoothing. At the same time, the underlying mathematical basis for understanding the mechanics of hotspot mapping is being developed using a new formulation called the lattice diagram.

Recently, epidemiologic base maps were created and used to calculate regional odds ratios on a set of 37 variously sized grids of the Cape Cod region, each representing a different spatial resolution from which to aggregate cases and controls. Crude odds ratios were calculated using data from an on-going population-based case-control study. A separate line of analysis, adaptive rate smoothing, is being used to aggregate cases and controls into target regions adaptively sized to contain a given number of controls. For this, a database of 2,438,282 subject residence pairs was created. Case/control odds for a series of residence years (1983, 1973, 1963, 1953 and 1943) and control smoothing parameters (n = 10, 25, 50, and 100) were then calculated. The resulting set of risk surfaces are now being analyzed.

Additionally, the functionality of the GIS was recently used to generate data for an airborne exposure source to test theoretical approaches to analyzing the kinds of small data sets typical of environmental epidemiological studies. A theoretical approach called the lattice diagram is currently being used to explore fundamental issues that plague environmental epidemiological studies, including exposure misclassification, ecologic bias and its mechanisms, as well as other issues.

The development of these technologies is significant because they provide promising new analytic methodologies for the field of environmental epidemiology. Significant to visualize potential relationships between geography, environment and cancer risk. These advances are practical in that they are producing a new cancer/environment mapping system for a defined population. At the same time, theoretical advances are being made by developing new basic tools to analyze and understand environmental epidemiological data.

For More Information Contact:

David M. Ozonoff
Boston University
Department of Environmental Health
715 Albany St., Talbot 4W
Boston, Massachusetts 02118
Phone: 617-638-4620

To receive monthly mailings of the Research Briefs, send your email address to