ArcUser Online

October - December 2007
Search ArcUser
ArcUser Main Current Issue Previous Issues Subscribe Advertise Submit An Article

E-mail to a Friend

Spatial Statistics Provide New Insights
Researcher sees possible links between MS and other diseases
By Susan Harp, Esri Writer


Spatial statistics have opened up a new way of looking at a debilitating disease that reportedly affects 400,000 people worldwide.

Normalized count of multiple sclerosis deaths by county for 1998 over 1990 census data

George de Mestral envisioned the design of the Velcro fastener in 1948 while picking burr-covered seedpods from his dog's fur after a mountain hike. As the story goes, the Swiss citizen stopped to observe the sticking qualities of Mother Nature's design and made the leap to a new, creative application. With the avalanche of information available to researchers today, the catalyst that helps produce this kind of "ah hah!" moment is extremely valuable.

For Megan M. Blewett, a young 21st-century researcher, spatial geography played a role in both her ah-hah! experience and her research. Blewett turned 18 in 2007, but five years ago, she was already reading a neuroscience textbook and asking questions about a mysterious disease—multiple sclerosis (MS)—that she found described in its pages. Blewett said, "I started researching MS when I was 12 and have since fallen in love with discovering the insights spatial statistics can give."

MS affects the central nervous system. Although its cause is unknown, many researchers think environmental triggers might be a factor. This unsolved puzzle caught Blewett's attention. She started collecting data about MS cases in her home state of New Jersey, learned to map their distribution with GIS, and has been using spatial statistics tools to analyze that distribution. She has continued reading about the neurological and biochemical aspects of the disease. However, her ah hah! moment occurred at a science fair while she was talking with one of the judges about her map of MS distribution in New Jersey.

"I just got lucky there," commented Blewett. "I was looking at a state map of MS distribution and saw that my county, Morris County, has a high incidence of MS. You could see individual towns, and I knew the town next to me had a high incidence of Lyme disease." A bacterial infection, Lyme disease is spread by tick-borne spirochetes. She was already using ArcGIS Desktop to map MS distribution, so when she started thinking about a possible Lyme disease correlation, she added Lyme data to her map layers.

click to enlarge
Normalized count of Lyme disease deaths by county for 1998 over 1990 census data

"I saw all these correlations and results that I hadn't been able to see before and still don't think I would have been able to see if I had been using more conventional chemical research to look at individual proteins at work," Blewett added. "Spatial statistics allowed me to see the bigger picture. Then I zoomed in to look at proteins at work in MS and related demyelinating diseases. I like to say my research path is analogous to reading the summary before reading the book."

The data collection process was one of the harder parts her research. Data came from TheDataWeb, an online set of libraries, and DataFerrett, a data mining tool, both provided free to the public by the United States Bureau of the Census and the Centers for Disease Control and Prevention (CDC). When Lyme disease data was not available online, Blewett had to contact the state epidemiologist and request data. Eventually she received data from every state. "To my knowledge, it is the largest standardized dataset of Lyme information in existence," saidBlewett about the dataset. She also said she is willing to make the data available to other researchers.

Blewett ran a correlation analysis. She calculated a Pearson's correlation coefficient (r) (for the normally distributed variables) or Kendall's tau-b or Spearman's rho for data that was not normally distributed. All correlation analyses assumed a linear relationship between the variables so the appropriate coefficient was calculated for pairs of variables in three datasets. All variable values were converted to z-scores for use in a regression analysis. Finally, cartographic analyses compared MS, Lyme (from other specified arthropod-borne diseases data), and control data from external cause of death data.

"The two disease distributions were pretty similar—they correlate and the control doesn't," explained Blewett. "Biochemically they are also very similar, so it has just taken off from there." She hypothesizes that both diseases may share a common spirochetal basis, and MS might develop from a secondary tick bite.

click to enlarge
Control data: normalized count of external causes of death by county for 1998 over 1990 census data

Blewett consulted with Esri spatial statistics expert Lauren Scott on using GIS in her research. "While biologists and medical researchers investigate this hypothesis at the cellular level, Megan's work examines the spatial fingerprint of these two diseases at broad spatial scales and then tests hypotheses regarding their spatial correlation," said Scott.

"I wish to expand my research from a national to a global scale, while also testing my models in smaller geographic areas," Blewett said. "A recent study suggests that MS is, in fact, 50 percent more common than previously predicted."

Blewett presented her work at the 2006 Esri International User Conference and participated in the Academic Fair during the 2006 Esri Health GIS Conference. In 2007, she was accepted into several top universities and awarded seventh place in the prestigious 66th Annual Intel Science Talent Search. For more information, contact Megan Blewett at or or visit

Contact Us | Privacy | Legal | Site Map