WASHINGTON — Whooping cough first sickened the Illinois high school cheerleaders, then it struck the football players, the cross-country team and the band.
As it spread within the Chicago suburb of McHenry County in late 2011, another outbreak took place — on social media. A small business called Sickweather said it detected the online flare-up on Twitter and Facebook postings in early October that year. That’s about two weeks before local health officials issued a public statement.
Now, U.S. agencies want to expand their use of social media to spot potential biological attacks and outbreaks of deadly infectious diseases, including the new H7N9 avian flu that has killed dozens of people in China.
“That’s the Holy Grail,” said Mark Dredze, an assistant research professor at Johns Hopkins University in Baltimore and a Sickweather adviser. “We’d love these systems to tell us there’s a brand new disease and it’s going to be a big thing.”
The online disease trackers have had mixed results, with academics criticizing a tool by Mountain View, Calif.-based Google, the world’s biggest Internet search engine, for overestimating the number of U.S. influenza cases in recent years.
The system, dubbed Google Flu Trends, relied on search terms. It was never intended to be used on its own, said Matt Mohebbi, a former company engineer who helped create the tool.
Kelly Mason, a Google spokeswoman, said the company is open to feedback on how it can refine Flu Trends to help estimate influenza levels and “complement existing surveillance systems.”
Companies such as Sickweather and Boston-based Epidemico are trying to get past the noise on the Internet. They rely on computer algorithms to scan social media and news articles for references to disease like “whooping cough.” They try to screen out unrelated posts that might use “sick” (when they mean cool or insane) or “Bieber fever” (obsessed with pop star Justin Bieber).
The work also involves humans, in case the filters don’t catch everything and the algorithms exaggerate illness reports.
“The big advantage of social media is you can get a lot more data, and you can get it more quickly and more economically,” said Henry Niman, a biomedical researcher and president of Pittsburgh-based Recombinomics, which analyzes viral evolution and the spread of disease. “It is a matter of fine-tuning that data so you come up with results that are more reliable.”
In McHenry County, the initial cases of whooping cough, or pertussis, were among the cheerleaders at Cary-Grove High School, according to local health officials.
By April 2012, 336 cases had been reported in the county, according to local public health data. The bacterial disease, known for its telltale “whoop” sound as infected people gasp for breath, can be fatal, especially in infants.
Sickweather now provides notices when it detects signs of illness through its website and free real-time alerts, said Graham Dodge, chief executive officer of the closely held company. It didn’t do so during the pertussis outbreak because the website was about to go live, he said.
“It was so early for us, we didn’t even know if it was accurate,” he said.
Dodge has met with U.S. agency officials and hopes to work with them after Sickweather raises more money for expansion.
Social media may help the Centers for Disease Control and Prevention, part of the U.S. Department of Health and Human Services, better spot outbreaks in smaller communities, said Matthew Biggerstaff, a CDC epidemiologist.
It held a contest this year in which 11 teams comprised of academics, and industry and health officials competed to develop the best way of using social media to predict the flu season, he said. The winner will be announced by June 20.
“These digital surveillance tools can help you start picking up signals a little bit earlier,” Biggerstaff said. “They give you access to real-time information before it starts registering on our system.”
CDC officials typically track flu by monitoring reports of doctor visits, hospitalizations and deaths, he said.
Those reports may take a while to trickle in. On Dec. 3, 2012, the agency announced the flu season was off to an early start. Sickweather’s Twitter post arrived about six weeks earlier: “Oh, hello #Flu, you’re a little early this year.”
In a separate effort, the health department sought guidance from businesses in February on how it could tap social media to keep tabs on H7N9 and the Middle East Respiratory Syndrome. The MERS virus has killed at least 86.
It’s looking for near real-time analysis of Twitter posts that might show developments in the two diseases. Such tools might have other applications, including alerting government officials when hospitals or nursing homes are evacuating patients in a disaster, according to a federal website.
“If there’s a reason for it, we can direct resources to a community in need,” Kelly Bennett, a public health analyst for a health department office, said in an interview.
U.S. officials want more than online chatter on the flu. A Homeland Security office monitors social media to detect naturally occurring health emergencies — and “acts of malice,” S.Y. Lee, an agency spokesman, said in an email. It is also seeking algorithms to analyze social media for reports of disease or biological attacks.
The disease trackers also follow local news online.
“Infectious disease may have taken weeks or months to be known,” to public health officials, Brownstein said. “Now it can take a week or two.”