Caroline Buckee uses mobile health data to track and predict the spread of deadly diseases around the world, and her new predictive models help officials know where to focus disease-fighting efforts.
Although U.S. residents worry considerably more about catching West Nile Virus than coming down with malaria, the Centers for Disease Control and Prevention estimates that there were 198 million worldwide cases of malaria claiming 500,000 people in 2013. This is mostly an ongoing problem for regions in Africa, and the CDC isn’t the only organization trying to stop the mosquito-borne disease.
Harvard epidemiologist Caroline Buckee was studying the genetics of malaria in Kenya in the mid-2000s while her then-husband Nathan Eagle was analyzing cell phone data for carriers. They realized they could combine their skills to track the spread of the parasite.
Cell phone data can actually help officials stop people from bringing deadly diseases to unaffected locations.
“As epidemiologists, we try to figure out how to prioritize intervention in different places,” Buckee said. “With cell phone data, we can follow people over time, build up a map of where everybody is, then use mathematical models to predict the spread of disease. This is the future of epidemiology.”
Information on public travel patterns is indispensable during an epidemic. It lets epidemiologists track how an outbreak is spreading and predict where it will go next, as well as how people’s movements are changing in response to an outbreak. Authorities can then figure out where to focus disease-fighting resources, how to manage travel restrictions and whether their efforts are working at all.
Historically, epidemiologists and public-health workers have had to resort to low-tech methods like interviews, surveys and theoretical models to find out how people are moving around. But these methods give unreliable and patchy data, Buckee says, and are especially difficult in undeveloped areas.
Cell phone location data is much more accurate. It tells where people are moving to and from, and even who they are coming into contact with. Best of all, it’s essentially instantaneous.
The cholera outbreak that followed the 2010 earthquake in Haiti offered a hint of how useful cell phone data could be in tracking and predicting movement patterns.
Swedish researchers analyzed movement data from two million phones, supplied by Haiti’s largest mobile carrier. Within 12 hours, they found that 630,000 people left Port-au-Prince within three weeks, often for the same places they had spent the Christmas and New Years’ holidays. They were, in essence, able to follow the outbreak.
These types of calculations can change the course of a disease.
Buckee studied how predictions based on cell phone data compared to real life in Pakistan, for example, where dengue fever has made alarming inroads.
Dengue is the quickest spreading mosquito-borne disease in the world, in part because of increasing travel, but also because climate change is helping the insect vector spread to new areas.
“Mosquitoes are now everywhere across the country,” Buckee explained. “They didn’t used to be.”
To assess the reliability of cell phone data predictions, she compared records of a large series of outbreaks across Pakistan in 2013 with a model based in part on mobility data from about 40 million phones. That mobile data predicted the spread and timing of epidemics accurately — and confirmed that people traveled much more than previous estimates.
The study estimated that between 2.4 million and 4.8 million phone subscribers moved between tehsils (roughly equivalent to counties) every day.
“We were surprised how much travel there was,” Buckee said. “There was no way to see that with the standard approach — and it’s just going to increase.”
The travel problem is an issue in malaria-affected regions as well.
In Kenya, Buckee said, travelers were spreading malaria, in some cases more so than mosquitoes.
Her team’s analysis of records from 15 million phones showed which settlements were especially likely to receive visitors from malaria-ridden areas. They found that people returning from malarial hotspots on the edge of Lake Victoria played an important role in transmitting the parasite.
Cell phone data can help even if it isn’t pulled from the country with the outbreak. For example, Buckee and other researchers weren’t able to secure operator data from affected countries during the 2014 Ebola outbreak in West Africa.
Instead, they used data from Kenya, Senegal and Cote d’Ivoire to predict how people in places like Guinea and Sierra Leone were spreading the deadly disease. Their findings suggested that porous national borders were a problem: Even if flights between large cities were suspended, people could still cross easily from country to country in rural areas.
Despite evidence that cell phone data is an effective means of studying the spread of deadly diseases, the method isn’t without challenges. Getting the data in the first place is one of the biggest hurdles.
Mobile data is strictly regulated and closely guarded by the telecom companies that collect them. Privacy is a major issue, especially in the developed world.
In the future, Buckee imagines a technological solution to the privacy problem: smartphone apps with opt-in features that share users’ location and medical symptoms. Making systems like these available to authorities and health-care workers could help them develop real-time epidemic response plans, including text message warnings to people in disease hotspots.
She is also fascinated with the idea of linking cell phone data to genomics to track how pathogens mutate over time and space, helping create better vaccines.
“The implications are across the board,” Buckee said. “As far as I’m concerned, more data is better.”