Northeastern University researchers use computers to simulate 20 million virtual Ebola outbreaks each week. Yale scientists are building three models that project the spread of the deadly disease. And a team at Boston Children’s Hospital is combing through data to gauge whether medical interventions are working.
These researchers may not be on West Africa’s front lines against Ebola, but they are close behind, an army of data scientists and epidemiologists often toiling around the clock to stop the disease.
Together they are providing a constant stream of evidence that is beginning to reveal the weak spots of the epidemic. For example, scientists’ models are beginning to identify basic patterns of who is being infected and when and how Ebola is being spread, which could help identify the most meaningful ways to intervene.
“It’s kind of all hands on deck right now,” said Caroline Buckee, an assistant professor of epidemiology at Harvard School of Public Health who is using mobility maps based on cellphone network data to project how people — and the deadly virus — will move around a region.
“The first priority has to be clinical, just getting people cared for and treated,” Buckee added. “But there’s a whole lot of planning and policy that has to happen: projections for how many beds they’ll need and how many gowns to send out, and where to send them.”
Northeastern researchers began, almost on a whim, to look into the West Africa Ebola outbreak in early August. Now, with their work continuing unabated, the computer scientists and physicists on the team laugh when asked how many hours a week they work, noting that the days blend together when they don’t distinguish weekdays from weekends and they arrive and leave work when it is dark outside.
They are driven by the knowledge their predictions could help stem the spread of the virus and the hope these insights could aid policy decisions or the efforts of doctors and nurses ministering to patients. In a study published in the journal Eurosurveillance last month , they showed that even drastically cutting back on air travel will, at most, delay the arrival of the virus by days or weeks to a given country.
“The motivation for all of us is you’re doing something that would be useful,” said Alessandro Vespignani, a professor of physics and computer science at Northeastern University. “Life for us, it’s easy — we change a line of code. Just think about the people who are traveling there.”
In a study published in the Annals of Internal Medicine last week, Yale researchers built a model of the disease’s spread using data from the outbreak of Ebola in Uganda 2000-2001 and from this August in Liberia. They found that people who ultimately died of Ebola infected more than two other people on average, while survivors infected fewer than one other person on average. According to their model, isolating three-quarters of the patients within the first four days that they show symptoms would help eliminate the disease.
“What we’re trying to do is to get a general understanding of who is mostly responsible for the transmission, try to predict what is the pattern going to be like — how many people are going to be infected,” said Dan Yamin, a postdoctoral researcher at Yale School of Public Health. “By that, we can find what should be the optimal strategy.”
Last Thursday, in a separate study published in Science , Yale researchers reported a prediction of another model that tested the effects of interventions, such as tracing all the contacts of a sick person. Their model suggests the virus will infect 224 new people each day in Liberia alone by December unless a variety of interventions are implemented. Unlike past outbreaks, where single measures such as ensuring burials were done in a sanitary way may have been sufficient, their study found that a combination would be necessary to contain the disease.
More than a week before the World Health Organization announced that Ebola had broken out in Guinea in March, a team of Boston Children’s Hospital researchers began to detect the disease in online news reports and social media. On March 14, their software — called HealthMap — registered the first sign of “a new disease that we do not know the name was reported in the prefecture of Macenta located 800 kilometers from Conakry, killing 8 people and leaving several others contaminated.”
Now, the researchers are harnessing the technology to understand how actions such as a public health education campaign in the community or delivery of protective gear are changing the spread of the disease on the ground.
“It’s hard to know if it’s correlation or causation, but it seems that interventions seem to precede decreases in transmission rates,” said Maia Majumder, a computational epidemiology research fellow at HealthMap.
Harvard’s Buckee works with other academics through the nonprofit Flowminder Foundation to use cellphone data to model how people are moving around. In high-income countries, those patterns are often defined by daily commutes, Buckee said, but in low-income countries, a daily path may be driven by agriculture, casual labor migration, and travel between rural and urban areas.
Buckee then overlays those mobility maps with more traditional approaches to understanding how a disease spreads to try to predict where it will show up next.
Much of the fear and attention in the United States is focused on the cases that may arrive from Africa, but disease modelers are far more concerned about the possibility the disease could arrive in a country with a less sophisticated health care system.
“It’s reasonable to expect more cases in the US,” Vespignani said. “Its not reasonable to expect a flood.”