Mapping Ebola

Published on December 27, 2014
Why cellphone data could be the key to tracking the next pandemic.

Eric Fischer uses cell phone data to visualize population movement in cities around the world. This series maps populations (London and Taipai) based on where they take photos, and whether they are locals (blue), tourists (red) or unkown (yellow).

The current Ebola outbreak has brought media attention to the public health practice known as ‘contact tracing.’ This practice – of locating every person who has come in contact with an infected individual – has proved to be valuable in managing contagious diseases, such as STDs and tuberculosis.

While tracing sexual contacts in order to manage the spread of sexually-transmitted diseases may be challenging, the job is significantly harder when tracking casual social contacts of patients with tuberculosis. Or Ebola.

According to the CDC, contact tracing can stop Ebola in its tracks. What it entails, as recent media attention on the topic has taught us, is finding everyone who comes in direct contact with a symptomatic Ebola patient. The more the patient travels, or the more crowded the locations, the more difficult the tracing process gets. While contacts may be easier to find at certain sites like hospitals, airplanes and cruise ships, it becomes virtually impossible to do so in densely crowded urban environments like buses, subways and restaurants. One missed contact, says the CDC, can keep spreading the epidemic. The likelihood of successful tracing in dense conglomerations, therefore, seems to very low. But it may not have to be.

Over the last decade cell phone penetration has grown exponentially with cell phone to person ratios reaching or exceeding 1 in most urban centers around the world. In other words, almost everyone has a cell phone, especially in cities. And our cellphones almost always know where we are. People often voluntarily share their whereabouts on social media like Facebook and Foursquare, by “checking in” at their destinations, or allow friends to know where they are by activating GPS location apps like “Find My Friends.” Apps like Trip Journal and Pathbook allow users to share their travel routes with friends and family, in real time. MIT Media Lab researchers recently demonstrated that cellphone usage creates fairly unique digital signatures: Knowing roughly where and when a user is four times a day is sufficient to uniquely identify them 95% of the time.

Now imagine such digital footprints being made available to public health epidemiologists tasked with the herculean goal of tracing everyone that came in contact with a contagious person. Suddenly, contact tracing can rely on triangulated geospatial data rather than on memory and recall.

Mobile phone communication data can be used to investigate structural properties of large-scale social networks and human mobility patterns. Here is an analysis of mobile phone communication networks. Here the nodes correspond to individuals and ties connect pairs of individuals who communicated with one another during the study period. Ties have been colored based on their strength, aggregate communication time over the study period, such that weak ties appear in yellow and strong ties appear in red. Structure and tie strengths in mobile communication networks by JP Onnela, J Saramäki, J Hyvönen, G Szabó, D Lazer, K Kaski, J Kertész, and AL Barabási. PNAS 104, 7332 (2007).

While many citizens may not be ready to turn over such private data to the government without substantial protective laws in place, anonymized data can still play a meaningful role. Researchers around the world are already using cellphone data to identify populations at risk. Using data from Digicel, Haiti’s largest mobile carrier, the Stockholm based Flowminder Foundation demonstrated that the majority of the 630,000 Haitians who left Port-au-Prince on the day of the earthquake had done so within three weeks, and generally went to the same places where they had spent Christmas or New Year’s Eve. Access to such anonymized data would have huge policy and programming implications for urban epidemiologists. Even if personal route maps are not accessed, predicting association and travel patterns from anonymized data holds tremendous promise for shaping response: Where do most people from a particular neighborhood go to work every morning? Do most travel by bus or train? Do patients discharged from emergency departments tend to go straight home or do they often go to the nearest pharmacy or the market? Introduce a layer of granularity and we can ask even more specific questions: How does the staff who took care of the sick patient usually go home? Who were the persons that were in the vicinity of the index patient?

The relatively safe asymptomatic phase when the Ebola patient is not contagious may have prevented the wide spread of disease in the United States – where known returnees from West Africa vigilantly self-monitor for signs of infection for a period of three weeks. But this may not be the case with the next outbreak when another pathogen may be even more contagious and incubation periods even shorter. In dense urban centers of Ebola affected countries with a large number of contagious symptomatic patients, contact tracing has been incredibly important, and remains a resource-intensive enterprise.

The relatively safe asymptomatic phase when the Ebola patient is not contagious may have prevented the wide spread of disease in the United States – where known returnees from West Africa vigilantly self-monitor for signs of infection for a period of three weeks. But this may not be the case with the next outbreak when another pathogen may be even more contagious and incubation periods even shorter.

Underlying the study of anonymized cell phone data are general principles of human behavior. Their adaptation requires some degree of contextual intelligence, but the principles are robust. While Big Data may help predict population trends, small granular data may be invaluable for contact tracing. Were epidemiologists indeed allowed access to everyone’s GPS trackers, they would potentially be able to identify contacts by studying the patient’s timeline, and the digital footprints of all that may have been present at the same locations. The result would be a sub-set that would include those most at risk. Such monitoring would also alert authorities to potential contacts boarding buses, trucks, trains, and flights.

Taipei, visualized by mapping where individuals took photos: locals (blue), tourists (red) or unknown (yellow).

With unbridled access, we do run the risk of information overload, a distinct characteristic of our times. It is also likely that incremental increases in the level of contact tracing are likely to yield diminishing benefits. To learn to use cellphone effectively, researchers need greater access than they currently have. Anonymized cellphone data, whether retrospective, or in real-time, will allow epidemiologists to learn to sift through hundreds of millions of data points, unmasking unforeseen limitations and unleashing the potential for vast public good.

Privacy concerns are legitimate, and in the post-Snowden era, there will be significant resistance to government-initiated monitoring of personal cellphone data. Last month, in the United States, the Florida Supreme Court ruled that ‘stingrays’ – simulated stealth cellphone towers used to track people’s movements – were in violation of the Fourth Amendment, and could not be deployed by law enforcement agencies without a warrant.

There is, of course, global precedence for limiting personal liberties for public safety. Quarantine, in the context of epidemic disease, is one example, deployed with varying levels of success over the centuries. Since 1985, the United Nations has laid out the Siracusa principles that provide officials with guidance on how to weigh the health and rights of communities against the health and rights of individuals.

In times of high crises citizens could be given a choice to allow their movements to be accessed so that public health responders can do their job better. And citizens may demand that granting access be incumbent on stringent laws that prevent data misuse. Our timely ability to strike an appropriate balance between our collective health and individual rights may be one of the critical factors that shape the outcome of this and other epidemics. Until then, depriving epidemiologists of these data renders them unable to harness the full power of today’s technology to protect our health.

Satchit Balsari is an assistant professor and chief of the Global Emergency Medicine Division at NewYork-Presbyterian's Weill Cornell emergency department. He is as visiting scientist at the Harvard School of Public Health.

Jennifer Leaning is Francois-Xavier Bagnoud Professor of Health and Human Rights at the Harvard School of Public Health and Director, FXB Center for Health and Human Rights.

Jukka-Pekka Onnela is assistant professor in the Dept. of Biostatistics at the Harvard School of Public Health.

Tarun Khanna is Jorge Paulo Lemann Professor at Harvard Business School and Director, Harvard South Asia Institute.

The authors have employed mobile technology to study disease surveillance and social homophily at the world’s largest mass gathering, the Kumbh Mela in India, attended by over 100 million people.

This article originally appeared in Issue 15 of Emergency Physicians International.

e.g. Global, Research, or India
e.g. Features, Opinion, or Research
Browse by Category
    Most Popular
      Download Latest Issue
      Issue 19