Virtual Plague

The carnage was breathtaking. over the course of the five-day epidemic, bodies in various states of decay piled up in city squares. Travelers unwittingly carried the virulent pathogen from rural villages to urban centers and back again. Through their efforts to help their fellows, first-responders on the scene unintentionally spread the highly contagious disease as they rushed from victim to victim. As the outbreak continued to rage out of control, social chaos ensued.

This doomsday scenario didn't happen in our world. It happened in the World of Warcraft, a massively multi-player online roleplaying game (MMORPG) to which some 9 million people subscribe. The uncontrolled plague that broke out in September 2005 was a virtual one, but the lessons that can be learned from it have real value, according to researchers at Tufts University's Initiative for the Forecasting and Modeling of Infectious Diseases (InForMID), an interdisciplinary collaboration of experts from the medical school and their colleagues around the world.

Games like WoW could prove to be a novel and powerful weapon in InForMID researchers' arsenal as they seek to model and predict outbreaks of infectious disease.

The Plague of '05

In their September 2007 article in the British medical journal The Lancet Infectious Diseases, research assistant Eric Lofgren and Nina Fefferman, Ph.D., codirector of InForMID and research assistant professor in the Department of Public Health and Family Medicine and assistant research professor at Rutgers University, report on the virtual virus and how online video games like WoW could serve as research models for real-world epidemics.

First released by Blizzard Entertainment in 1994, the World of Warcraft game franchise has been around since the dawn of the Internet. Players of the game create characters known as avatars that can make friends and fight enemies in the virtual world. Through their adventures and quests, players accrue power, prestige and health-- or lose it all in one epic battle.

Intentionally introduced into the game by its designers, the infection-- known as "Corrupted Blood"--was meant only to challenge high-ranking players as they battled a powerful winged serpent named "Hakkar." But several factors that parallel real-life outbreak scenarios led to the pathogen's uncontrolled spread through the World of Warcraft, Lofgren and Fefferman note in their article. Although only high-ranking players could enter the infectious area of the game to battle Hakkar, they exposed susceptible lower-ranking players to the disease upon their return to busy urban centers populated by players at all levels. In a scenario akin to Europeans bringing smallpox to the New World, many of these weaker players died almost instantly.

Other true-to-life aspects of the game perpetuated the disease. As in real life, many WoW players own pets. The game penalizes players whose pets die, so players often dismiss their pets from battle scenes until the threat is averted. In the case of Corrupted Blood, some players unwittingly dismissed infected pets.When they called them back--often into bustling urban environments--the pets triggered new outbreaks of the disease. According to Lofgren and Fefferman, pets were the major means of transmission, much like the livestock and vermin that have been implicated in plague and influenza outbreaks throughout human history.

To Lofgren and Fefferman, the mimetic elements of World of Warcraft, intentionally programmed by the game's designers, make it a perfect laboratory in which to study infectious disease. Epidemiologists could program a bug that would act like HIV or influenza when unleashed on a virtual world. But the behavior of the very real people playing the game is the unpredictable element that makes WoW such a novel and attractive way to model what really goes on during epidemics.

Most models of disease, says Fefferman, fail to account for social behavior in any detail because it's so hard to study. It would be unethical--not to mention an expensive logistical nightmare--to induce an outbreak or run controlled studies on people during naturally occurring plague times. Warcraft, however, offers a completely safe means of observing the range of human responses during epidemics. "That's the beauty of Warcraft. It models social behavior," says Fefferman."Most epidemiologists presuppose human behavior, or collectively throw up their hands."

In the case of Corrupted Blood, some players endowed with healing abilities (like real life EMTs or ER docs) attempted to revive their fallen comrades. Their altruistic behavior wound up perpetuating the spread of the disease by sending infected characters back into the game to interact with other characters. By contrast, some players who became infected seem to have intentionally spread the disease out of spite or curiosity.

"Most of the time, when there is a threat to society, bonding together helps," Fefferman notes. "But in the case of infectious disease, the threat comes from society. The threat and society are the same thing."

Model Citizens

Of course, Fefferman acknowledges, she and her colleagues have no idea if people respond to infectious outbreaks online the same way that they would in real life. Some may act out socially unacceptable fantasies online; some may take risks with an avatar they wouldn't take with their own lives--what Fefferman dubbed "the stupid factor." But these differences between real life and e-life can be quantified. Fefferman cites work under way at MIT's Media Lab, where researcher Dan Ariely studies "eRationality," how people behave and make decisions in electronic environments, and Judith Donath investigates identity and society in the online world.

Based on their work and her own experience, Fefferman suspects that most players probably do behave in WoW much as they would in the real world. Many players maintain the same avatar for years--and may develop strong relationships with other online characters. (Fefferman, who doesn't consider herself a real "gamer," has had a WoW character since before the game had graphics, in the earliest days of the Internet.) It's this inclusion of spontaneous human behavior that could make WoW an incredibly useful tool for epidemiologists.

"Prediction has a long history," says Elena Naumova, associate professor of public health and family medicine and founder and director of InForMID."But the explosion of computing has allowed us to design analytical tools that are more flexible, more responsive,more sophisticated than ever." To study infectious diseases like influenza, for example, epidemiologists try to quantify several basic characteristics about the pathogen: the probability of transmission from one host to another; the rate of recovery; whether the pathogen confers immunity once the host recovers and relative susceptibility among individuals. Armed with this data, epidemiologists build two types of models, statistical and mathematical, which work in concert to describe an epidemic. Naumova, a biostatistician, specializes in statistical models, which start with data observed in the real world.With enough information, statistical models can predict the number of new cases of an infectious disease over time--say, the number of cases of the flu in January versus the number in May. Statistical models provide answers, usually in the form of numbers.

Meanwhile, Fefferman makes mathematical models, which she calls the "exact complement" to statistical models."Mathematical models give you a why, a logical story that you tell yourself."As a mathematical ecologist, Fefferman belongs to a small but growing academic field concerned with the "why" questions about evolution. A math major as an undergrad at Princeton, Fefferman, G05, was also interested in biology and worked on her Ph.D. in ecology at Tufts under J.Michael Reed, professor of biology in the School of Arts and Sciences. Her work in conservation ecology brought her into Naumova's lab, where she has been a research assistant professor in family medicine and public health since 2005.

Beyond the Hive

Although the Corrupted Blood epidemic garnered more press than some real-world diseases do, scoring mentions in The Economist, Time and on the BBC, the paper was actually just one small project on Fefferman's to-do list. She is specifically interested in trying to understand how social complexity arises, particularly with respect to infectious disease. "I don't particularly care whether I am studying termites, or humans, or zebras or Tribbles," she says--though so far, she's focused mainly on termites, bees and wasps, a class of insects that have taken social complexity to extremes.

Among these social insects, individuals have prescribed roles: The queen lays eggs; the workers tend to her; the soldiers defend the nest. The individual is subsumed by the hive, which becomes something of a super-organism. Like an engineer designing a machine, Fefferman builds models that predict how the insects should behave to maximize the hive's output. Then she tests the accuracy of her models by checking them against what really happens among colonies of insects in the lab.

Her work has implications beyond the hive. Last year, she collaborated with Doug Brugge, associate professor of public health and family medicine at Tufts, to model the impact of urban development on crime and health in Boston's Chinatown neighborhood.And in her capacity as a research professor at Rutgers' Center for Dynamic Data Analysis (DyDAn), Fefferman, in partnership with the Department of Homeland Security, is building tools to help the government better assess its public health surveillance data.

Through institutes like the Centers for Disease Control, the federal government monitors reports of 14 diseases, ranging from the highly infectious smallpox virus to the highly lethal hanta virus. The mathematical tools Fefferman creates quantify the randomness in the numbers, helping identify when a spike in public health data truly indicates an outbreak. These may seem like disparate projects, but to Fefferman, each of her collaborations is just another opportunity to study social animals and infectious disease. "That's why I love InForMID," she says.

InForMID began as an informal network of researchers from various fields who shared an interest in forecasting diseases. In 2005, Naumova secured a $6 million, five-year grant from the National Institutes of Health to formalize the group, which today includes about 50 researchers around the world, from the Dana-Farber Cancer Institute in Boston to the National University of Singapore. The experts--specialists from fields like applied mathematics, ecology, public health and urban planning--all bring different perspectives to the study of infectious disease, says Fefferman. Their collaboration could help solve some enduring epidemiological mysteries, like why certain illnesses occur more during specific seasons of the year.

Looking for Patterns

Take salmonella poisoning, a food-borne illness that peaks during the summer months in North America. That's when food is more likely to be left unrefrigerated on picnic tables and beach blankets, allowing the Salmonella enteroca bacteria responsible for the ailment to multiply. However, as globalization brings more produce to American grocery stores from tropical climates where the bacteria thrive year-round, the seasonal pattern of infection has flattened out. Salmonella poisoning is still the result of human error in food handling, says Naumova, but epidemiologists must account for how globalization adds complexity to the picture.

"It's human nature that we are always looking for patterns so we can avoid mishaps. These are the deep roots of epidemiology," Naumova points out. "Our models will never tell us exactly when an outbreak will hit, but they will give us clues which may give us some lead time" to respond.

That lead time may be especially important with influenza viruses, which--like salmonella-- exhibit seasonality. Unlike salmonella, however, researchers can't say why. The popular notion that flu hits in the winter months because people spend more time cooped up together doesn't have adequate research to back it up, according to a paper Lofgren, Fefferman, Naumova and colleagues published in the Journal of Virology last summer. Nor do hypotheses about how El Nino, or light/dark cycles or seasonal changes in the human diet affect influenza seasonality.

"There are a bunch of theories, none good enough to explain everything," Fefferman admits. "What would happen if you pulled an immunologist and a virologist and get them to work on the same paper? It would be cool to get that perspective." Bringing those experts together is exactly the goal of InForMID, says Naumova, who adds that the interdisciplinary research generates not just novel answers, but novel questions to pursue.

Jacqueline Mitchell is a senior health sciences writer in Tufts' Office of Publications. She can be reached at jacqueline.mitchell@tufts.edu