Just months after the first cases of swine flu appeared in April 2009, millions of Americans had gotten sick and some had even died. By the end of the year, the virus had spread worldwide, creating the first influenza pandemic since 1968.
As drug companies produced a vaccine that prevented millions more from catching this flu, researchers participating in an international project called MIDAS were simulating disease spread. The simulations let them explore how the pandemic might unfold, who was more likely to get sick and which interventions might protect the most people. The results helped inform public policy decisions.
Flu and You
To create the pandemic flu simulations, the MIDAS researchers use computer models to build virtual cities, countries and even continents. Here, thousands of pretend people go to school, work, stores and other places. The researchers base the residents' activities on information about actual people like you.
Stephen Eubank, a physicist at Virginia Tech University in Blacksburg and part of the MIDAS team, has modeled virtual versions of major U.S. metropolitan areas using local transportation and census data. In Eubank's cities, there really are six (or fewer) degrees of separation between any two people—making it easy for germs to spread.
"Viruses don't care much about geography," says Eubank. "They care about social networks and how people come into contact with each other."
Another key part of studying the spread of infection with computers involves developing a virtual version of the germ. To model its spread as realistically as possible, the researchers track down everything known about the infectious agent. Eubank, who has studied plague, smallpox and anthrax, has gathered information on how each agent spreads between people, how contagious it is and how long it takes for an infected person to show symptoms.
When they don't know the actual characteristics of an infectious disease, the MIDAS researchers use health reports and scientific data collected during earlier outbreaks to estimate what a future one might be like.
During graduate school, Christina Mills modeled a pandemic flu before we had ever heard of swine flu (also called H1N1). She did a lot of research scouring the shelves for scientific articles that discussed the 1918 Spanish flu—a pandemic that killed between 20 and 40 million people worldwide.
"It was very old-fashioned," says Mills, who was studying international health at Harvard Medical School in Boston, Massachusetts. "I couldn't just type a search word into Google™ and get the necessary information." The hunt eventually led her to the 1918 transmission rates.
With all the modeling pieces in place, the MIDAS researchers invite policymakers to ask questions that can be answered using the models. Questions range from What happens if we don't do anything? to How many people could be protected if we intervene?
The researchers create different simulations that change the variables, like the contagiousness of the virus or the number of people taking "snow days"—Eubank's term for people who voluntarily hang out at home to avoid infection.
"What's so great about the computer simulations is that you can try out different situations that you can't create in real societies," says Eubank.
With more than 250 possible combinations to simulate, Eubank says he relies on statisticians to help him determine which arrangements will produce the most informative results.
"It's easy to come up with questions," says Mills. "The hard part is figuring out which ones we should—and could—answer."
Because of the amount of data and calculations involved, the simulations run on high-performance computers that can simulate a 180-day outbreak in a matter of hours. Eubank uses software programs to take snapshots of the pretend pandemic as it occurs.
"I know exactly when a virtual person gets infected, shows symptoms and recovers," says Eubank, explaining that the computer records every change in disease state.
Eubank and other researchers modeling 2009 H1N1 pandemic flu have simulated outbreak scenarios in communities across the United States. The results suggested that early vaccination of school kids best reduced disease spread, while vaccinating elders became more important later on. The simulations also indicated that people at risk for serious complications—like pregnant women or individuals with pre-existing health problems—should be given antiviral medicines to take at the first signs of illness.
While the results generated by the simulations are useful, Eubank stresses that they're not a guarantee of what actually will happen. He and others often will ask different models the same questions and, when the models agree, they'll have more confidence in the predictions.
Meet the Simulators
Stephen Eubank started out studying high-energy physics but then got into modeling the dynamics of nonlinear systems, which are systems that can't be solved by adding up all of the parts. He has developed computational models to study natural languages, traffic patterns and financial markets. He plans to use the infectious disease models to study how behaviors, like smoking, spread through society.
Christina Mills has a Sc.D. (like a Ph.D.) and an M.D. For her, modeling infectious diseases is a dream job because it combines her interests in math, biology and human health. While most of her colleagues with double degrees practice "bench-to-bedside" research in which they translate lab findings into patient care, Mills says she'll stick with the "computers-to-clinics" approach.