Can Social Media Help Contain Ebola?

Can Social Media Help Contain Ebola?

Patrick Sawyer, Nigeria’s first Ebola patient, collapsed at the international airport in Lagos on July 20. This Wednesday, more than six weeks later, the World Health Organization (WHO) said that it was monitoring at least 200 Nigerians for infection related to Sawyer’s case. Sawyer, a Liberian-American who had traveled from Monrovia, had carried the often-fatal disease to Africa’s most populous country, hundreds of miles from its origin. It was as if he had slipped through a crowd.

Fortunately for the people of Nigeria, crowds leave traces, even when the individuals within them disappear. As Ebola spreads, some epidemiologists are beginning to analyze those traces to guess where outbreaks might occur. They’re not only gathering data from diseased neighborhoods and hospitals. They’re also using sources like flight data, Twitter mentions, and cellphone location services to track the disease from afar. Researchers, in short, are sifting through the detritus of mobile lives to map the spread of an unprecedented outbreak.

Some of the results have been surprising in their accuracy. Northeastern University’s Alessandro Vespignani used flight records and population data to produce models predicting that Ebola might spread to Senegal; that was confirmed by Senegalese authorities on August 29. HealthMap, a web project that aggregates and analyzes information from news sites, social media, and other resources, noticed the outbreak over a week before the WHO’s public announcement. Flowminder, a Stockholm-based nonprofit, turned anonymized cellphone location data into a map of West African transportation trends. The group plans eventually to chart “the most connected communities in different countries, given different scenarios of Ebola appearance,” Caroline Buckee, a Harvard epidemiologist who sits on Flowminder’s board, told Foreign Policy.

These projects are cutting-edge, but they’re far from the fringe. Independent network scientists often send their findings to the WHO, which uses it, along with locally collected data, to support national health ministries. “We are part of a group of systems,” John Brownstein, HealthMap’s co-founder, told Foreign Policy, “that provide intelligence-gathering to the WHO so they can do their work.”

HealthMap’s Ebola map, Sept. 4. 

Models such as those created by Flowminder provide one way to predict how the disease might spread, but they come with a crucial catch: Many of the patterns uncovered in mobility data are historical, not real-time. And it’s not clear how those patterns might change in response to crisis. Previous tendencies in regional travel, after all, don’t say much about how people modify their behavior during an epidemic. Commuters to Monrovia might cancel trips as the disease spreads. “That is the classic paradox of any model that has a social component,” said Vespignani. New conditions generate new behavior. And though Buckee says that real-time analysis is “theoretically possible,” there’s no evidence that any such analyses are currently in place. And who’s to say that people with cellphones are the same people who are carrying the disease?

More importantly, remotely collected data can’t stand on its own. The accuracy of any large-scale forecast depends on granular detail — the kind that’s usually collected through the dangerous, demanding work of sending health workers through individual hospitals and neighborhoods, where they could themselves be infected with the disease. To accurately know how quickly a virus will spread, for example, you need to know its transmission rate. “Modeling layers are coupled,” Northeastern’s Vespignani said. “Having poor data in one layer affects the other layers.”

It’s partly for this reason that the WHO remains focused on “maintaining the flow of information on real cases” instead of on remotely collected data, Christopher Dye, a director of strategy at the WHO, told Foreign Policy. “The central task for us,” Dye said, “is to keep track of the number of cases and to make sure there isn’t underreporting.” That means information-gathering on the ground.

But full reporting is a long way off. In Monrovia, a city of about 1 million, hospitals are overflowing, quarantined residents are slipping back into the general population, and nurses have gone on strike because they lack protective equipment. More than half of all cases in urban areas might be unreported. Models have outrun the information they need. Uncertainty fills the gap.