Thematic Maps

Mapping Disease (1 & 2)

Although John Snow was able to use his maps to prove a cause and effect, he was restricted to what he could illustrate by hand and by the limited data he had available. His beautiful hand crafted map of central London included the 500 deaths near the Broad Street Pump, but he may have wished that he could map all of the more than 10,000 deaths that occurred in Britain that year. Or to be more comprehensive, he may have additionally mapped the 53,000 deaths from epidemics in 1848-49 and the 23,000 British deaths from 1831-32. Certainly a comparison of the geographic patterns in cholera deaths from epidemic to epidemic would have provided valuable insights into transmission of the disease.

It was not until the advent of geographic information systems (GIS), more than 100 years later, however, along with significant improvements in data collection and quality, that epidemiologists were finally able to simultaneously visually represent enormous datasets on disease as well as possible sources of risk.

Developed in the mid 1960s by scientists in the Canadian Government, GIS, a computerized system for the manipulation, analysis and display of geographically referenced data, has enabled innovative research on a wide range of environmental health issues. Researchers, for example, are now able to map multi-million record datasets on, pesticides and air pollutants, among others, and to relate them to data on health issues such as asthma, low birth weight or cancer (English et al. 1999; Ritz and Yu 1999; Krautheim and Aldrich 1997; Reynolds et al. 2002 ).

While GIS presents an enormous opportunity for epidemiologists and others, widespread access to GIS technology and data also presents challenges in ensuring scientific accuracy and cartographic standards. In the words of Mark Monmonier, author of How to Lie With Maps, “because of advances in low-cost computer graphics, inadvertent yet serious cartographic lies can appear respectable and accurate.” He further points out that “wary map users must watch out for statistical maps carefully contrived to prove the points of self-promoting scientists, manipulating politicians, misleading advertisers and other propagandists” (Monmonier 1991).

One source of deception, for example, lies in a map’s greatest strength — and arguably it’s greatest weakness — it’s ability to distill often complex sets of data into simple, sometimes simplistic, representations. Rarely, for example, do mapmakers adequately represent the underlying uncertainty in data. In most cases, maps represent a mean value (e.g., predicted disease rates, average soil permeability), and fail to represent standard errors or other measures of uncertainty. Maps further mislead by often “discretizing” otherwise continuous processes. A map may represent a clear demarcation of cancer rates, for example, from county to county where a clear demarcation does not exist in the real world.

Finally, relying on sometimes inaccurate visual representations of natural phenomena can also lead to erroneous conclusions about associations or cause and effect. Scientific pursuits involving GIS often rely simply on visualization rather than empirical methods to validate perceived patterns and associations. The movement beyond simply mapping to empirically validating spatial concepts and “broadening the analytical toolbox” in GIS has been slow (Anselin 2000).

Until very recently GIS software packages have failed to incorporate spatial statistical functionality. In a recent article, researchers at the Universities of Michigan and Illinois, for example, argue that “most GIS packages still offer little in the way of statistical analysis and state-of-the-art spatial statistical methods” (Bao et al. 2000).

The recent developments of the S-PLUS extension for ArcView, the ArcGIS Geostatistical Analyst extension and the announcement that ESRI, the leading developer in GIS software and SAS, makers of statistical software, are planning joint software developments suggests that the barrier between statistical analysis and GIS is diminishing, though it still remains.

Nevertheless, the statistical limitations of GIS haven’t precluded interesting and sophisticated spatial analysis. Spatial statistics is an important data analysis tool and has helped epidemiologists explore disease patterns and refine traditional statistical techniques.

© 2006 ZevRoss Spatial Analysis