Public health lessons from virtual game worlds

It’s challenging to model disease spread during epidemics. Simple mathematical models such as the “general epidemic” model make assumptions about constant population size, homogeneous mixing, and constant recovery rates, but can only go so far in predicting an outbreak’s severity (See “Mathematical Modeling of Epidemics”). They typically don’t consider the role of animals, the impact of human behavior, and variable population susceptibility–i.e., immunosuppressed individuals are more susceptible to getting a disease than those with normal immune function. (See my January column, “The Growing Number of Immunocompromised.”)

Models are often used to assess the impact of isolation and quarantine on disease spread. (See “Mathematical Models of Isolation and Quarantine.”) An important variable to consider when contemplating the institution of quarantine and/or isolation is “R ₀,” or “attack rate,” which represents the infectiousness of individuals who have the disease. R ₀ estimates the average number of people a sick person will infect. When R ₀ is greater than 1, an epidemic is likely to occur. Obviously then, the goal is to get R ₀ to less than 1.

But because epidemics aren’t usually recognized until they are either well established or almost over (take for example, influenza, which can infect people before they develop symptoms), epidemiologists typically use historical data on epidemics that occurred days, months, or even years ago when creating models. An important new tool in observing epidemics in real time has emerged in online virtual games. In September 2005, Blizzard Entertainment unleashed a highly contagious and virulent plague called “Corrupted Blood” in its virtual world, World of Warcraft.

The epidemic quickly spread to densely populated cities that led to widespread death and social chaos. In the September 2007 Lancet Infect Dis journal article “The Untapped Potential of Virtual Game Worlds to Shed Light on Real World Epidemics,” E. T. Lofgren and N. H. Fefferman estimated that the R ₀ for the epidemic was 100 per hour for the cities, which is astonishingly high–not even influenza, which is highly contagious, has an R ₀ (around 2) that high. In World of Warcraft, people are able to transport themselves to different locations quickly, which also helped the disease spread. While these variables would be extreme in a real-world plague, the virtual epidemic did simulate many aspects of an actual emerging infectious disease outbreak. For example, “Corrupted Blood” was a zoonotic disease transmissible between animals and humans. Many of the most lethal newly emerging infectious diseases (H5N1 avian influenza, SARS, and Nipah virus) are zoonotic.

Fefferman came to Princeton University in November 2007 to discuss her research and findings in the virtual epidemic. World of Warcraft is one of the largest virtual world communities with millions of players from diverse backgrounds and geographic locations. Many of these players are highly committed and form close online relationships. As the epidemic spread, Blizzard Entertainment implemented a quarantine and made “public-health” announcements–all to no avail. They ultimately rebooted the system to eliminate the plague.

Despite the virtual nature of the epidemic, important insights were gained from the event. For example, participants exhibited behaviors such as altruism, fear, and curiosity. Many came online just to see what was going on before quickly logging off. Unfortunately, the outbreak, which lasted five days, was over before Lofgren and Fefferman could get access to the real-time data. Regardless, they were able to demonstrate the usefulness of studying and modeling similar virtual events in the future.

Policy makers and public-health officials should take note of this event as a potential strategy for modeling and understanding how diseases spread. Millions of people play in these virtual worlds, and while most of their characters, or “avatars,” don’t permanently die, they have strong reasons to avoid getting sick and dying. Their behavior during a virtual epidemic would provide an important window into human responses to these types of crises and give epidemiologists and public-health preparedness professionals a powerful preparedness exercise.

To conduct these types of virtual pandemic-preparedness activities, there should be public-private partnerships between public-health officials and virtual reality companies. There are partnerships being developed between games creators and scientists, defense analysts, and disaster-response specialists who build games as educational exercises for professionals such as firefighters and HAZMAT teams. (See “Games Get Serious.”) These efforts could be expanded to develop virtual epidemics in online fantasy worlds with known parameters such as R ₀, modes of transmission, and degrees of virulence. Public-health officials and epidemic modelers could also issue announcements and other interventions to see how they affect human behavior and disease spread. The Centers for Disease Control and Prevention could facilitate these joint ventures.

Yes, virtual game worlds are fantasy, but they could be important tools in improving our understanding of how humans behave during deadly infectious disease epidemics.