Researchers Can Follow Outbreaks in Real Time
By Edwin L. Aguirre
Fever. Cough. Sore throat. Runny nose. Body aches and headache. Fatigue.
These are some of the classic symptoms of the flu, a highly contagious respiratory illness caused by influenza viruses. In the United States, the flu season usually begins in October and can last until May, with the peak in January or February.
A team of researchers from UMass Lowell’s Computer Science Department, Harvard Medical School’s Department of Population Medicine and Scientific Systems Co. Inc. is now using online social networks such as Twitter and Facebook to help improve the prediction of influenza levels within a population and keep track of its spread.
Called the Social Network-Enabled Flu Trends, or SNEFT, the system uses a continuous data-collection framework that monitors all flu-related tweets. The team’s research is supported in part with a $200,000 grant from the National Institutes of Health.
“Studies have shown that preventive measures can be taken to contain the outbreak, provided early detection can be made,” says computer science Assoc. Prof. Benyuan Liu, who is a member of the team.
Liu says the Centers for Disease Control and Prevention (CDC) monitors cases of influenza-like illnesses, but since the diagnoses are made and reported by doctors manually, there could be a one- to two-week delay between the time a patient is examined and the moment that data becomes available to the CDC.
“Public health authorities need to be forewarned at the earliest, to ensure an effective preventive intervention,” he says.
Seasonal influenza epidemics result in about 3 million to 5 million cases of severe illness and about 250,000 to 500,000 deaths worldwide each year, says Liu.
“In 1918, the so-called ‘Spanish flu’ killed an estimated 20 (million) to 40 million people worldwide, and since then, the influenza virus has mutated to a variety of particularly virulent forms like SARS and H1N1, against which no prior immunity exists,” he says.
Enhancing Public Health Preparedness
Twitter, a micro-blogging service, has become a popular platform for people to update their daily activities, resulting in billions of pieces of information being posted and shared on the web.
UMass Lowell doctoral student and team member Harshavardhan Achrekar tapped into Twitter and extracted 9.2 million influenza-related user posts from 2009 to 2011, taking advantage of Twitter’s ability to provide an almost-instantaneous snapshot of current epidemic conditions and building comprehensive mathematical models that can estimate nationwide flu activity.
“We consider Twitter users within the United States as ‘sensors’ and the collective message exchanges they post describing their flu symptoms as early indicators and robust predictors of flu activities,” says Liu.
Results presented in scientific publications show that these posts on Twitter closely match the number of flu-like cases reported by the CDC.
“By applying novel filtering techniques, we can build models that accurately predict seasonal influenza levels across different regions within the United States and possibly among different age groups,” says Liu. “Thus, the online social network’s ability to track the flu in real time provides an opportunity to significantly enhance public health preparedness against influenza epidemics and other pandemics.”