Researchers Use Facebook to Model H1N1 Disease Spread
"When did you first learn about the swine flu outbreak? Have you searched the Internet for additional information on the swine flu outbreak? If a vaccine for swine flu became available, would you want to be vaccinated?"
These questions first appeared on Facebook on Saturday, April 25, just a day or two after concerns of H1N1, or swine flu, swept across the country.
"I didn’t want to miss the opportunity to gather data on early reactions to the outbreak," said Lauren Ancel Meyers, a mathematical biologist at the University of Texas at Austin who models disease spread and posted the questions.
Meyers and her collaborator, Alison Galvani at Yale University, are newcomers to a National Institutes of Health research program called the Models of Infectious Disease Agent Study (MIDAS) that develops computational models of how infectious diseases emerge, spread, and can be contained. The results can help public health officials plan for and perhaps even prevent contagious outbreaks.
Key to the researchers' MIDAS modeling project is surveying people on how they perceive health risks. The researchers will use this information to build a dynamic model that simulates how changes in decision-making influence patterns of disease spread. The model will help them and others identify the strategies that improve adherence to interventions and reduce the spread of disease.
"Modelers have made great strides in building detailed and predictive models, but one of the important, missing ingredients is how individual-level decisions impact disease dynamics and how disease dynamics, in turn, change people's behavior," Meyers explained.
Ordinarily, disease modelers operate in a virtual world where a hypothetical infectious agent starts spreading across a city, state, country, or continent. The models are built on data collected from previous outbreaks and actual populations, but they still must assume a lot of "what ifs."
Such as: What if the bug is really infectious and deadly? What if it's resistant to the drugs we have? What if we have a vaccine that's effective, but there's only enough of it to immunize part of the population? What if we close international airports? What if people don't adhere to the recommended interventions?
These are some of the questions MIDAS modelers have been asking about potential diseases like pandemic flu, which could emerge and infect people across the globe.
"As soon as we and the rest of the world learned about the newly emerged flu strain," Meyers said, "we realized that this could provide an invaluable window into how people respond to actual outbreaks."
But crafting a survey that returns scientifically significant results takes time, Meyers said. You have to ask the right questions in the right way and make sure that a demographically diverse group of people respond. To capture data from the earliest days of the H1N1 outbreak, Meyers, Galvani, and their team quickly drafted around 20 questions and posted them on Facebook. Within the first few days, they received 50 responses.
"Some of them were my friends," said Meyers, who joined the online community just six months ago at the urging of her younger sister. "But some were not, and I thought, 'I wish I knew more about these people.'"
The preliminary results helped this new MIDAS team refine the questions, develop new ones, and identify a representative cross-section of the population to contact. Two days after launching the Facebook questions, they worked with a professional survey company to distribute questionnaires to a broad range of people. By April 28, the survey company had collected 265 responses. Today, it continues to collect information. The original questions, still posted on Facebook, receive occasional replies.
Meyers says that the group is interested in tracking how people's answers change as public health officials issue new information or guidance about H1N1. The differences could reflect what happens to people's perceptions, behaviors, and choices as a disease outbreak evolves. Incorporating this information, added Meyers, will truly advance the field of disease modeling.