Mining Search Data After Big Events Could Improve Public Health Response

Jun 7, 2017

If you’re like millions of other Americans, when a big event happens –a  shooting, a disease outbreak, a contentious election –you scour the internet to make sense of what’s going on.

Those millions upon millions of searches offer a wealth of data for researchers trying to figure out just what people want to know in the minutes, hours and days following significant events. Now, a team of researchers analyzing internet use habits following the 2012 shooting at Sandy Hook Elementary concluded that search can be used to help public health officials more effectively disseminate information during times of crisis.

"It gives you a sense of the public psyche around an issue," said Nir Menachemi, the lead researcher and a professor at the School of Public Health at Indiana University-Purdue University Indianapolis. "When we have a Zika outbreak or situation, you literally can look at how seriously people in areas are taking this issue. You can see what they’re clicking on.”

Menachemi and his team analyzed more than 5.6 million Yahoo search queries made in the two weeks before and the two weeks after the school shooting at Sandy Hook in Connecticut.

The researchers found a more than 50 percent increase in queries related to guns in the two weeks after the shooting than during the two weeks before the shooting. Similarly, searches for “shooting incident,” “ammunition” and “gun-related laws” were all searched for much more frequently in the two weeks following the shooting.

Special access to anonymized data from Yahoo allowed the researchers to see what websites users visited after they searched for gun-related words. In the past, research has mostly focused on what people were searching for, not what websites they were visiting after.

Knowing what people click on, and when, could help public health officials understand how the public reacts to outbreaks like Zika virus, what protective measures they take themselves and even people’s spending habits during the crisis.

“You can see whether or not it leads them to buy certain protective things like [products with] DEET, or other anti-mosquito products,” Menachemi said. “Then you can reposition your public health response.”

And that response often lags far behind events happening on the ground. To start, public health officials could do better at leveraging social media to get word out about real-time public health emergencies, Menachemi said. Application of his research could help public health officials understand what the public wants and needs to know more immediately.

“There are probably strategies that can be developed to better figure out more modern ways of interacting with folks,” he said. “There are good examples of this, but I don’t see it happening on a wide scale.”

Menachemi and his team could literally follow people from what they searched for to where they ended up, and where they often ended up was a surprise to them: retail gun websites. Of the 5.6 million searches researchers reviewed, more than 1.7 million of them ended up visiting a website where guns are available for purchase.

“The main reason people typically buy guns is for protection,” Menachemi says. An event like the Sandy Hook school shooting “sort of prompts people to think about their self-protection.”

Indeed, many people who were searching for information in the aftermath of the Sandy Hook school shooting wound up at either pro-gun or gun control sites, Menachemi said. That, he added, led to a missed opportunity for local government or other community institutions to provide more neutral information about the shooting.

“Getting one side of information is always bad,” he said. But those sites – pro-gun or gun control advocacy groups – are often updated more regularly than government sites, and they’re search engine optimized, meaning they show up higher in search engine results than sites with more neutral information – and are more likely to get clicked on. “So they’re getting crowded out,” Menachemi said.

Researchers elsewhere have used search engine data to estimate how effective a flu vaccine is in particular areas, or if searching for the term “anorexia” following media reports of celebrity eating disorders leads people to develop the disorder themselves. Researchers often also mine social media data from Facebook and Twitter to spot potential public health issues like who is getting sick, where they are, and what illness they are contracting.

Better understanding an individual’s search behaviors – anonymously, Menachemi is quick to add – could help officials better understand how their web-surfing behaviors change over time, and that sequence of events could shed more insight into better tailoring a public health response in times of crisis.

“We’re bound only by our imagination in terms of what application such analysis can be useful for in a given situation,” he said. “It gets people thinking.”