The first reports started coming in by mobile phone. Concerned users in Syria opened up the website of HumanitarianTracker.org and filled out a basic web form. Someone was sick; it looked like polio. Taha Kass-Hout, a co-founder of the platform, recalls how the spotty initial data trickled in. Concerned, he reached out to the Syrian opposition’s aid body — the Assistance Coordination Unit. Soon, they were sending him nightly spreadsheets of reported cases. He and his team manually pulled the figures and updated Humanitarian Tracker’s open source database. Polio, a nearly eradicated disease, had returned to Syria.
Within months, it was clear that crowdsourced reports such as this had noticed something that traditional epidemic surveillance systems had missed — or at least vastly underestimated. The World Health Organization reported 36 cases of polio in 2013-14. Researchers writing in the Annals of Global Health found 46 more using nontraditional surveillance methods, including data collected by Humanitarian Tracker.
The key to spotting the outbreak was what Kass-Hout calls the data “mosaic effect.” WHO polio surveillance relies heavily on government data from confirmed cases of a disease. But those monitoring networks have blind spots: where health systems have collapsed, where states are weak, where medical testing centers are lacking or where politics gets in the way.
Over the last half decade, researchers have begun to demonstrate that the gaps can be filled by adding in a variety of data sources, from local news to user-generated reports. “Every piece is telling you something about the bigger picture,” Kass-Hout said.
Mosaic data could transform epidemic response, providing real-time information about where a disease is spreading, who’s affected, and what control methods are most effective. But for now, that revolution remains only tentative. Even as data tools have proliferated, they haven’t yet been linked to policymaking except in isolated cases. Governments, organizations, and public health agencies are grappling with how to connect information with action.
The data evolution
Information is among the most vital weapons public health systems have during epidemics. But throughout history, such data has also been limited by surveillance methods, testing capabilities, and often, politics.
During the SARS epidemic of 2002, for example, China was accused of withholding data that would have indicated the severity of the infection, for fear of economic fallout. The resulting information vacuum delayed an international response and may have exacerbated the overall toll of the outbreak.
Tech firms were among the first to see the potential of online sources to supplement deficient surveillance networks, and even avoid politicized data. Sick people often turn to the internet for help, searching for “flu” or posting about it on social media; Google.org began looking at ways to use that data more than a decade ago.
One of the tools that emerged was Riff, an open source data platform that allowed organizations to “separate signal from noise” when monitoring an assortment of sources. Created by a Google.org-funded startup, the Innovative Support to Emergencies, Diseases and Disasters, Riff was deployed first during the swine flu outbreak of 2009, and then during the devastating earthquake that hit Haiti in 2010.
“We paired the need with the resources” using Riff, recalls Kass-Hout, who helped develop the tool. “Somebody [on the ground] can say there is a water sanitation issue, and immediately via text message, we can identify groups that were able to help.”
By the start of this decade, a number of agencies and organizations were looking for ways to exploit big data analytics, including the U.S. Centers for Disease Control and Prevention and the U.S. military, as well as researchers at Boston Children’s Hospital who built HealthMap and at the University of Oslo who built DHIS.
But it was the Ebola epidemic in 2014 that elevated data monitoring tools from a niche corner of public health to center stage. The epidemic surprised public health systems, which had never seen an Ebola outbreak larger than a few hundred people. The primary countries affected — Liberia, Sierra Leone and Guinea — lacked government surveillance to effectively track the spread of the disease.
Responding organizations literally didn’t know where to go. At UNICEF, field officers in Sierra Leone turned to their organization’s innovation team. “We need to know where people are,” ventures lead Christopher Fabian recalls them saying. His group within UNICEF began using a text-messaging system called U-Report that asked users for information about their communities and offered information about epidemic response in reply.
“The thing we know about epidemics is that they move fast and they move big, and if we are really trying to deal with a global health crisis the most important tool we have is information,” he said. “During Ebola, if we had had these systems [such as U-Report] in place [before the crisis], we may have been able to know better where to set up our resources, because we work with a finite set of resources.”
As data sources have grown, so too has their potential to inform policy. Ebola provided an obvious example: If organizations and governments had known where and how quickly the disease was spreading, they could have positioned resources and communication strategies accordingly.
Messaging is among the most powerful applications of epidemic data, particularly for new or emerging diseases. Educating communities not to touch Ebola patients or wash corpses, for example, was vital to stemming the spread of the disease; that messaging could have gone out more quickly and in a more targeted way with better data about who was affected.
Organizations and governments could also use data sources to better position and allocate resources, something that Kass-Hout says he saw start to happen during the Haiti earthquake. A number of NGOs shared information using the Riff platform, which allowed them to coordinate deployments and link them directly to user-generated reports of need.
Once those basic applications are checked off, policymakers could take the data a step further with the growing sophistication of mathematical modeling. The CDC, for example, used its BioMosaic platform to help predict and track Ebola infections in the United States by following diaspora populations’ movements in and out of West Africa.
“We also can pair the information we are using with other types of information to build risk projections,” said John Brownstein, chief information officer of Boston Children’s Hospital and director of HealthMap. “When you pair on mathematical modeling, you can look at what various mitigation strategies might be able to scorch out an epidemic, or to use transport data and the risk of movement to project where a virus could spread.”
Organizations and governments are now expanding their informational arsenals for tackling epidemics. But for now, both the data and the policy are playing catch-up to the speed of disease spread.
“The challenge for surveillance as a whole is turning it into action,” said Brownstein. HealthMap, for example, produces a large amount of high quality, open-source data, yet the data team knows relatively little about if and how it is used.
“One of the big challenges we face as HealthMap is that we push out a lot of information but we don’t get a lot back” in terms of how organizations are utilizing it to inform policy.
The success of data mapping projects may, in fact, be one of the reasons that policymakers have been slower to translate information into action. The proliferation of sources means there are few unified databases — a fact that can overwhelm both consumers and producers of the data. Phone users in West Africa during Ebola, for example, often received multiple text messages per day requesting information about their communities’ health situation, Kass-Hout recalls. “If 20 different organizations are asking you to send data, there is a lot of confusion,” he said. “One group is trying to push one software, other groups push another.”
Still, there are examples of cooperation. Some 30 governments, mostly in sub-Saharan Africa and South Asia, have adopted DHIS, a platform developed at the University of Oslo, to track disease outbreaks. Mobile provider Orange — one of the largest mobile networks in those regions — is now looking for ways that they can help augment that data by providing metrics about user mobility, said Nicolas de Cordes, vice president of marketing anticipation.
UNICEF Innovation also offers a model; its data tools are crafted with specific field applications in mind. The U-Report tool, for example, spotted rampant sexual exploitation in schools in Liberia after the Ebola outbreak ended; UNICEF leveraged that data to help put a response system into place. “If it doesn’t link back into action, it’s useless,” Fabian said. “We are already drowning in reports.”
The Zika virus offers an immediate test case for the growing importance of using data to craft policy. A few basic mitigation techniques can help slow the spread of the disease, for example limiting standing water in homes. Women in affected areas are also being advised to put off pregnancy. Governments and responding organizations could target their messaging with precision using big data to pinpoint the most affected communities.
UNICEF, which is now partnering with tech firms to link mobility and weather data with epidemic responses, expects the learning curve to get steeper and shorter. With Zika, “we are very much at a preliminary stage,” Fabian said.
“If you look at the arc, it takes a little bit of time to figure out how best to use these systems. We know that having these data sources allows us to be more powerful and effective; we now have a good set of partnerships including with Google, IBM, and Amadeus, and we have requests from our country officers. All this comes together to mean that in the very near future, [we can] be much more efficient,” he said.
Devex Impact associate editor Adva Saldinger contributed reporting.
With potential to change the trajectory of crises, such as famines or the spread of diseases, the innovative use of data will drive a new era for global development. Throughout this monthlong Data Driven discussion, Devex and partners will explore how the data revolution is changing our approach to achieving development outcomes and reshaping the future of our industry. Help us drive the conversation forward by tagging #DataDriven and @devex.