Scale Issues in Spatially Referenced Health Data

The appropriate spatial scale will vary with the frequency of the disease or health event being analyzed. It is far easier to collect data at a fine scale and aggregate upward than it is to collect data at a large scale and then be forced to infer rates at a smaller scale. Because most health data are available only at a high level of spatial aggregation (e.g., county, zip code, or census tract level) and a great deal of within-unit spatial variation is typically present in data attributes, full spatial analysis is not feasible. For example, the National Cancer Atlas9 allows queries by state, county, or State Economic Area based on cancer location and occurrence interval. Although there is substantial geographic variation in cancer mortality within states and within counties, this is not reflected in National Cancer Atlas data. Cancer registries in the United States release data only at the zip code level, and because zip codes are arbitrary units with no inherent geographic or geological significance, they are inferior to census tracts or census block groups for demonstrating spatial variation and drawing conclusions with respect to social and economic variations in health disparities (e.g., Krieger et al., 2002). Although staff members at individual cancer registries and some researchers may—under very restrictive conditions—be able to gain access to spatially more specific locations of patient residences, such access is highly variable and depends on study protocols and Institutional Review Board10 (IRB) restrictions.

Data Access Issues

Why are data so fragmentary and why is it so difficult to obtain data at a fine scale? The fragmentation of data has its roots in government structure, with responsibilities for data collection divided among local health departments, state health departments, and the federal CDC. For conditions—such as cancers—that do not need to be reported to the CDC via local and state health departments, reporting is to local cancer registries. Similarly, trauma cases are reported voluntarily to local trauma registries. Consequently, there is no central repository of health data in the United States and there is considerable variation in the formats and location requirements of the data that are reported.

The reason that the location of incident cases is so difficult to obtain




IRBs are groups established by individual institutions (universities, private companies, etc.) with the charge to review research to assure the protection of the rights and welfare of the human subjects involved in biomedical research.

The National Academies of Sciences, Engineering, and Medicine
500 Fifth St. N.W. | Washington, D.C. 20001

Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement