Mapping the Next Epidemic
If you were the chief public health officer in Kansas City and the country were in the midst of a virulent flu epidemic, one thing you would want to know is where the greatest danger of further contagion might come from. Infected truck drivers driving through on I-70? Airline passengers from major population hubs such as New York or Miami? Small-town residents from rural counties, where the caseload so far was low, leaving the population more vulnerable to infection?
These are the types of questions which might be addressed by a sophisticated epidemiological mapping program developed under the direction of Howard Vollum Adjunct Professor of Science Richard Crandall ’69 at Reed’s Center for Advanced Computation, and biology professor Steve Arch. Crandall’s research was supported by two grants from the Defense Advanced Research Projects Agency (DARPA). Participants in the research included Elizabeth Cahill ’03, and Laura Rude ’01, as well Multnomah County epidemiologist Amy Sullivan.
The software program, USA-Flu v1.5, uses powerful mathematical tools to model the spread of seasonal influenza in the U.S. in a typical flu season (generally, mid-fall to mid-spring). The model drills down in extremely fine detail to predict infection and recovery rates at the level of each county in the U.S., using data from the Census Bureau and the Centers for Disease Control. It thus combines a high level of demographic specificity with comprehensive geographical breadth, compared with traditional epidemiological mapping tools, according to Crandall. “Every pixel [on the map] is 10 square kilometers— about four square miles, the size of Lake Oswego,” he explains. “No one has that depth; it’s a deep computation.”
Although the model was developed for conventional influenza, which kills approximately 35,000 Americans in a typical year, it can be adapted to other infectious diseases, including avian flu and smallpox.
The mapping tool allows manipulation of several parameters in a given epidemic. These include vaccination rates among infants, children, the middle-aged and elderly; mortality rates; and the effectiveness of the vaccine (if any). Adjusting these factors can be instructive, showing, for instance, how a higher vaccination rate among children can lower fatalities among the elderly. “If you inoculate all the infants and children, the ‘baby-sneezes-on-grandpa’s-lap’ phenomenon has been removed,” Crandall explains. “This isn’t to save the children’s lives—they don’t die from flu. It’s to save the grandparents, because the communicability from children is incredible.”
Alun Lloyd, a mathematical epidemiologist at North Carolina State University, says that similar models are already in use in assessing bioterror threats, and that this specific software tool is potentially useful as well. “It could, perhaps, be used to give some idea of the distribution of outbreak sizes, or the places where outbreaks are most likely to occur or to have their major impact,” Lloyd says. “The model can be used to assess the relative effectiveness of different control strategies.”
Crandall’s group at Reed has also created a model that sketches HIV spread in Africa in coming years; additional research funding, says Crandall, would allow the group to refine the program with more accurate epidemiological data from across the continent.
The flu and HIV mapping programs, which are Mac-compatible, are available for download at http://academic.reed.edu/epi/tools.html.