Imagine an avian-influenza pandemic infecting a nationwide population of 281 million people over the course of 180 days. Given today's worldwide connectivity, supercomputers can simulate just that.
Dr. Ira Longini, a biostatistician in the Public Health Sciences Division, and a team of scientists from Los Alamos National Laboratory, have used supercomputers at the California lab to develop a model that predicts the possible course of a bird-flu pandemic. Their research, which appears in the April 11 issue of the Proceedings of the National Academy of Science, examines the nationwide spread of a pandemic-influenza virus strain, such as an evolved avian H5N1 virus, should it become transmissible between humans. The simulation rolls out a city- and census-tract-level picture that examines the impact of interventions — from antiviral therapy to school closures and travel restrictions — while the vaccine industry works to catch up with the evolving virus.
Mitigation strategies
"We believe that a large stockpile of avian influenza-based vaccine containing potential pandemic-influenza antigens, coupled with the capacity to rapidly make a better-matched vaccine based on human strains, would be the best strategy to mitigate pandemic influenza," said Longini and co-authors Drs. Timothy Germann, Kai Kadau and Catherine Macken of Los Alamos.
Their collaboration is supported by grants from the Department of Homeland Security and the National Institute of General Medical Sciences MIDAS (Models of Infectious Disease Agent Study) program.
"It's probably not going to be practical to contain a potential pandemic by merely trying to limit contact between people (such as by travel restrictions, quarantine or even closing schools), but we find that these measures are useful in buying time to produce and distribute sufficient quantities of vaccine and antiviral drugs," Germann said.
"Based on our results, combinations of mitigation strategies such as stockpiling vaccines or antiviral agents, along with social-distancing measures could be particularly effective in slowing pandemic flu spread in the United States," Longini said.
The results show that advance preparation of a modestly effective vaccine in large quantities appears to be preferable to waiting for the development of a well-matched vaccine that may not become available until a pandemic has already reached the United States.
"Because it is currently impossible to predict which of the diverging strains of avian H5N1 influenza virus is most likely to adapt to human transmission, studies of broadly cross-reactive avian-influenza-based vaccines with even modest immunogenicity in humans are important," Macken said. Ideally, both vaccine strategies would be done together — stockpile a modestly effective vaccine to use while the better-matched one is being developed, the authors suggest.
How it all computes
The computer simulation models a synthetic population that matches U.S. census demographics and worker-mobility data by randomly assigning the simulated individuals to households, workplaces, schools and the like. Department of Transportation travel data is used to model long-distance trips during the course of the simulation, realistically capturing the spread of the pandemic virus by airplane and other passenger travel across the United States. "In the highly mobile U.S. population, travel restrictions alone will not be enough to stop the spread; a mixture of many mitigation strategies is more likely to be effective than a few strictly enforced ones," Kadau said.
The model of disease transmission involves probabilities that any two people in a community will meet on any given day in any one of a number of settings, such as home or workplace. Thus, simulated disease transmission is more likely for two people in the same household and less likely for two people who have less in common. "So we are only computing the probability of any person becoming infected on any given day, and a roll of the dice is needed to decide whether they are infected or not," Germann said.
Power and possibilities
Other elements of randomness modify the simulated disease course. A significant fraction of infected people (33 percent in the present model) never develops clinical symptoms, although they are themselves infectious. In addition, the durations of the incubation and infectious periods can vary and are randomly chosen from distribution functions for each individual, involving more throws of the virtual dice.
"Computer models serve as virtual laboratories where researchers can study how infectious diseases might spread and what intervention strategies may lessen the impact of a real outbreak," said Jeremy M. Berg, director of the National Institute of General Medical Sciences. "This new work exemplifies the power of such models and could aid policy-makers and health officials as they plan for a possible future pandemic."
The pandemic-simulation model runs on a supercomputer known as Pink, a system software research platform in Los Alamos' Computer and Computational Science Division.
