Cell phones track our every move. A privacy advocate might see this as a problem. But an epidemiologist like Dr. Allison Aiello of the University of North Carolina sees it as an opportunity.
Aiello, along with researchers from across the country, recently published a paper showing that data from cell phones can be harnessed to identify people who are likely to contract the flu -- giving epidemiologists a potentially game-changing new tool in the fight against a disease that kills thousands of Americans every year.
"We believe this serves as a window into the future of epidemiological data collection," Aiello told The Huffington Post. Her colleagues agree.
"I think that the development of these kinds of models for medical data, where we can now collect data more frequently or in a more personalized manner, [for example by] using cell phones, is extremely important," Duke statistician Katherine Heller, a co-author on the paper, said in an email.
During the 2013 flu season, the study's authors gave 103 undergrads living in six University of Michigan dorms cell phones loaded with an app that used Bluetooth to record when they came into contact with one another. The undergrads also filled out a weekly online survey to report health-related behaviors, social interactions and flu symptoms. Anyone who reported flu symptoms was then tested for the flu virus.
This allowed the researchers to identify who contracted the flu after coming in contact with someone else in the cohort with the flu -- and what differentiated those people from others who did not contract the flu after such contact. They also monitored how long it took for people infected with the flu to recover.
They then used statistical analysis to develop an algorithm that predicts an individual's chance of becoming sick after exposure to the virus, based on behaviors such as sleep, drinking and smoking and personal traits such as sex, age and vaccination status.
The specifics of the algorithm aren't particularly surprising. People who drink and smoke are more susceptible to the flu and take longer to recover, while the opposite is true of those with good sleep and exercise habits. What makes the tool unusual, though, is that it takes all these factors into account at once to figure out each individual's likelihood of getting sick on any particular day.
Other researchers had previously built similar models, but they relied people self-reporting whom they'd interacted with -- which is highly prone to error -- rather than the direct observations that the cell phone data permitted.
The group studied by the researchers was too small and homogenous for the algorithm to be useful on a broad scale. But the researchers said they hoped this study would pave the way for more research that could lead to an algorithm that could predict anyone's chances of getting the flu -- allowing people to take measures that will keep them from getting sicker.