Release Date: December 27, 2002
BUFFALO, N.Y. -- A new method with the potential to quickly detect suspicious patterns in reported illnesses in specific geographic regions is being developed by a geographer at the University at Buffalo.
Combining cluster analysis with quality-control techniques traditionally used on assembly lines in factories, the method takes a novel approach to the problem of detecting potentially significant increases in the incidence of disease within specific geographic areas.
The method also allows for the testing of multiple hypotheses, an important facet when, as usually is the case, the cause for a jump in the incidence of disease is not known immediately.
To develop the method, Peter Rogerson, Ph.D., professor of geography at the University at Buffalo, is studying data supplied by Harvard Vanguard Medical Associates, which operates 14 managed-care clinics in the Boston, Mass., area.
Rogerson is using data on how many people with lower respiratory infections came into the clinics from January 1996 to October 1999 and then correlating that with information about small census tract areas where patients live.
"We are trying to model what occurred compared to a baseline and to see if there was any jump on one particular day or over a few days," he said.
The goal is to use the data to develop a model system with which to validate specific thresholds that could then be used in a new syndromic surveillance system.
"It's like tallying what you observed minus what you expected," said Rogerson. "In this case, we're looking to see where the incidence of respiratory infections jumped. We're doing this in order to figure out how to set the thresholds. If we set them too low, there will be too many false alarms; if we set them too high, we will miss things."
According to Rogerson, this type of analysis is similar to the quality-control function in factories where the primary role of engineers and technicians is to discover deviations from the norm.
"In factories, quality-control engineers measure the specifications of products that come off of assembly lines and compare them with what was expected," said Rogerson.
In his research, Rogerson uses the data about deviations to develop maps that reveal the geographic patterns of clusters of disease.
While other research efforts may involve the development of a graph based on a single county, for example, Rogerson's method develops a profile of symptoms within a whole series of counties in a region.
"Whether they stem from bioterrorism or from public health events with other causes, diseases don't limit themselves to the confines of particular counties or cities," said Rogerson. "We are able to broaden the geographic scale while getting better resolution, providing a more accurate and realistic picture of what's going on.
"I'm trying to pinpoint exactly when there is an increase in clusters of symptoms among people who live in all these counties," he explained.
Rogerson recently discussed his work, which is supported in part by a grant from the National Cancer Institute, at a conference on syndromic surveillance sponsored by the New York Academy of Medicine.