Malaria, which kills a population twice the size of Pittsburgh every year, has three components -- the parasite that causes it, the mosquitoes that carry it and the people who get it.
Public health officials often know the worst areas for the disease, because they can measure how many people in a region are infected. And they can identify risky areas by seeing how many mosquitoes and breeding areas there are.
But there's a vital third piece of information that has been harder to come by -- how often do people who are infected with malaria interact with those who aren't?
In today's issue of the journal Science, a research team led by Amy Wesolowski of Carnegie Mellon University has found some answers, thanks to the mountain of data generated by cell phone users.
By working with the major cell phone provider in Kenya, the team was able to track the movements of nearly 15 million people for a year, and compare that information with malaria infection rates.
The study not only provided more detailed information on the movement of Kenya's population than ever before, but suggested new ways to attack the malaria epidemic.
As expected, the team found that there was a lot of travel between outlying areas of the east African nation and Nairobi, the south central capital of 3 million residents. But to their surprise, they found that people living in the most heavily infected region, around Lake Victoria on the western border, also traveled frequently to surrounding areas with lower infection rates.
Ms. Wesolowski, a Ph.D. student in CMU's department of engineering and public policy, said the findings argue for tackling malaria in the high-infection area first, which is the opposite of what many African nations and nongovernmental organizations like to do.
Malaria control officials often like to focus on low-infection areas because it costs less and they have a better chance of scoring a statistical win against the disease. But fighting malaria in high-incidence areas might be better, she said, because it will have "larger spillover effects in the rest of the country."
Eventually, officials also might be able to send text messages to cell phone users who travel from high-infection areas to vulnerable regions, she said.
Unlike the flu, malaria doesn't spread from face-to-face contact, said Caroline Buckee, a Harvard School of Public Health professor who is also an author of the study. Instead, mosquitoes can bite an infected person and spread the parasite to scores of others.
So it's not just healthy people who need to use mosquito nets at night, she said, but infected ones, and the text messages could remind them of that.
The cell phone data also showed a lot more total migration throughout Kenya than the researchers had expected.
David Zarembka, a peacemaking coordinator who has lived in the Lake Victoria area of Kenya for 50 years, suggested several reasons for that.
First, he said, "part of the culture of Kenya is keeping up with your family and neighbors in person. When my wife was sick recently, all kinds of people came to visit her each day, including people from fairly far away."
Second, people in Kenya and other developing nations will often travel to other areas to work and then return home about once a month, which is why, on the last weekend of the month, "it's almost impossible to get a place on public transportation."
Finally, he said, governments and companies often regularly rotate employees to different parts of the country to cut down on the chance for corruption by reducing the odds that people will work in areas where they have lots of relatives and friends.
For all those reasons, cell phone data may give officials a much more useful picture of how to fight the spread of infectious diseases than any other source of information.
"These new electronic devices are probably the best way to track where diseases are moving," said Amesh Adalja, an infectious disease expert with the Center for Biosecurity of UPMC, who was not part of the new study.
Besides tracing people's movements, a spike in cell phone messages about illness can also signal the beginning of an epidemic, he said. In a similar way, Google and Twitter have used surges in certain keywords and search requests as a way of tracking the spread of diseases.
"You really have to understand how a disease is being spread in your geographic area to design the best control strategy," he said, and cell phone data "is almost like a heat signature that rescue workers use to look for someone."
The study, which also involved scientists from the University of Florida, Johns Hopkins University in Baltimore, the National Institutes of Health and the University of Oxford in England and Kenya, is part of a growing trend known as mHealth, which uses cell phones not only for research, but for diagnosis and treatment of diseases.
In Africa alone, Ms. Buckee said, the number of cell phone users is expected to have grown from 280 million four years ago to 735 million by the end of this year, and cuts across all demographic groups.
Researchers are experimenting with using cell phones for everything from communicating with health care workers in the field, to ordering supplies, to equipping them with software that can analyze blood samples.
The cell phone information is just beginning to be exploited, Ms. Wesolowski said. "Companies are collecting all this data, but they don't quite know what to do with it, and researchers are sitting around saying 'Hey, we need data.' "
Mark Roth: firstname.lastname@example.org or 412-263-1130.