Putting the brakes on a pandemic influenza outbreak
By Barry Whyte
Graphics by Aimee Drysdale
Related story: From mathematical models to public policy
Related story: MIDAS: The strengths of a network
Most experts agree that it is only a matter of time before a human flu pandemic grips the world. A novel flu strain that can transmit easily between humans could trigger a disease pandemic that would kill tens of millions of people worldwide. Researchers at the Virginia Bioinformatics Institute at Virginia Tech are using high-performance computer simulations to show how diseases like flu spread through large populations. They want to provide policymakers, decision-makers, and public health experts with a way to test health interventions that would help to put the brakes on the spread of a deadly influenza outbreak.
The stakes are high if an influenza pandemic were to take place. In a 2006 article in The Lancet, health experts who looked at data from the great 1918 flu pandemic predicted that 50-80 million could die worldwide depending on the virulence of the virus. The World Bank has concluded that a flu pandemic could cost the world economy a minimum of US$ 800 billion, an estimate that assumes a 2 percent loss in global gross domestic product. The combined effects of international transport networks and stretched health-care systems mean that a local flu outbreak could quickly escalate into a global crisis.
Researchers at the Network Dynamics and Simulation Science Laboratory (NDSSL) at the Virginia Bioinformatics Institute (VBI) are using computer simulations of mathematical models to investigate how infectious diseases like influenza emerge and spread through large populations of people. These simulations allow experts to test the impact of different public health interventions on such a spread of infectious agents.
Policymakers and other experts need to know in advance the optimal combination of measures to stop the spread of a disease outbreak. They need to know how effective it will be, for example, to isolate those infected with the virus and how to optimize the use of limited resources, like antiviral treatments. Powerful computer simulations provide valuable lessons and allow for better-informed decisions before a real outbreak takes place.
“Computer simulations that include the response of populations to a disease outbreak can help to estimate how an outbreak might spread and how different interventions may help to alleviate the disease burden.” — Stephen Eubank
“The global epidemic of avian influenza in bird populations, as well as the risk of a virulent form of the bird flu virus being transferred to humans, has helped to increase awareness of the very real threat posed by an influenza pandemic,” says Stephen Eubank, deputy director of the NDSSL. “Countries around the globe want to be as ready as possible if a flu pandemic were to be unleashed on the world. Our research team tries to help in this forward planning by creating virtual cities, populating them with synthetic people, and looking at the impact of social contacts on the spread of a disease pandemic in an urban environment. Computer simulations that include the response of populations to a disease outbreak can help to estimate how an outbreak might spread and how different interventions may help to alleviate the disease burden. This approach requires managing large amounts of data, and to cope with this challenge we use high-performance computing methods.”
Eubank and colleagues at VBI have been working as part of the MIDAS (Models of Infectious Disease Agent Study) group, a network that comprises principal investigators, scientific collaborators, software engineers, and data and computer experts, as well as students from research and informatics groups across the globe. The goal of MIDAS is to bring together modelers, policymakers, and the public health community to set priorities and design studies of infectious disease outbreaks. This involves making sure that MIDAS computer software is translated into useful tools for the public health community. It also encompasses sharing results and resources with the MIDAS network, policymakers, public health officials, and the scientific community. Researchers in MIDAS use a national consortium of supercomputers to run their computer programs. This includes the National Science Foundation-sponsored TeraGrid, which — with more than 750 teraflops of computing capacity and more than 30 petabytes of data storage — ranks as one of the world’s largest distributed cyberinfrastructures for open scientific research.
Modeling flu in the United States
“Perhaps one of the best ways to visualize the work of MIDAS is to look at a set of simulations for the city of Chicago that we recently completed for the United States Department of Health and Human Services,” says Eubank. “For the study, a high-performance cluster of 100 ‘blades,’ with four processors each, spent about 24 hours modeling the spread of pandemic influenza through the city. A blade is a modular unit of powerful computer hardware that is designed for easy integration into rack computer systems. The simulations run on these machines showed that the behavior of individuals can have a very large impact on how quickly a disease will spread through a city like Chicago, which has a population of more than 8.6 million people.”
To model Chicago at the level of detail needed, the researchers typically start with census information, public surveys, and transportation data. This gives them a realistic picture of the daily activities of the synthetic people built into the models and leads to detailed estimates of social contacts in an urban environment. A picture is built up of how social mixing patterns change under different interventions, such as closing schools or workplaces. When schools close, for example, young students require a caregiver’s attention. That can disrupt social mixing patterns at work if a working parent stays home, or make closing schools pointless if the children are placed in large day-care settings. This type of detailed individual behavior was built into the models.
On top of the social network, the researchers add important information on the influenza virus, such as how fast it spreads and how deadly it is. The researchers use the model to suggest the best mix of intervention strategies in a variety of scenarios, taking factors like these into account.
Says Eubank, “For Chicago, we were able to determine how a pandemic influenza outbreak might travel through the city. All of the simulations suggested that the combination of providing households with antiviral treatments in advance and minimizing social contact could play a major role in reducing the spread of illness. The timely initiation of these interventions and school closure were important contributory factors to putting the brakes on a pandemic.”
Learn more at: “Estimation of potential global pandemic influenza mortality on the basis of vital registry data from the 1918-20 pandemic: a quantitative analysis” (from The Lancet).
