Digital Data Helps Public Health Departments Keep Tabs on, Tackle the Flu

No one needs a reminder that this flu season was one of the worst in recent years. Preliminary data from the Centers for Disease Control suggests it may have been worse than the 2009 season, when the swine flu pandemic strain swept the country.

The good news, however, is that digital data is helping researchers and public health organizations provide more timely predictions about the spread of the flu and assessments of seasons in progress. The advantage is that public health officials no longer have to wait for a Monday-morning quarterback approach to flu season, but can stay on top of its spread as it happens.

“Traditional data put out by the CDC and others often has a delay of a week or two,” says Spencer Fox, who is pursuing a doctorate at the University of Texas in Austin and published research last fall on a computational model that explains why pandemics come at the end of flu season. “With digital data, you can get a sense of what’s going on with seasonal flu in real time, so you’re predicting one or two weeks ahead of traditional data sources.”

With real-time info, public health officials can help allocate the appropriate resources where and when they are necessary, as opposed to using their best guesses, which can change the outcome of flu season dramatically for the thousands affected each year.

SIGN UP: Get more news from the HealthTech newsletter in your inbox every two weeks

The Digital Data Toolbox Makes Flu Tracking Possible

Twitter, Wikipedia and Google Trends data are three of the most common types of digital information being used in Fox’s projects — people tweeting about being sick, searching flu-related pages on Wikipedia and Googling flu-related topics.

Fox says electronic medical records — though there are confidentiality issues to be worked out — and smart thermometers, such as the Nokia Thermo, which tracks temperatures and symptoms in apps connected to individual user profiles, pose other sources of useful data.

This flu season, researchers at the University of Iowa found they could make predictions up to three weeks earlier by using smart thermometers to track patients with fevers.

Meanwhile, researchers at the University of Chicago Medical Center analyzed insurance claims, demographic data and 1.7 billion geolocated Twitter messages to study the flu and people’s movements. They found the flu tends to originate in warm, southern parts of the United States and it spreads northward.

“Many of the interventions that public health officials run are in anticipation of flu season. So, if they know flu season is going to start early, they can start their public health awareness campaigns earlier. They can try to get people vaccinated and aware before the season happens,” Fox explains. “If they know it’s going to be a large flu season, they can allocate their resources accordingly.”

He foresees sequencing flu variants as delivering the most promising new data for trying to predict the size of an epidemic, but it will require an improved statistical framework for doing so.

The Springfield-Greene County Health Department in Missouri has been using a data visualization tool, LiveStories, which helps users provide context to their data with interactive charts, videos and benchmarks against peers.

“Public health has a long history of using data for data-based decision-making. … But what we’ve really struggled with is communicating that data effectively to the public,” says Kathryn Wall, the agency’s public health information administrator. “We can give incident rates, we can talk about all sorts of big, complicated statistical things, but we’ve come to recognize that to really effectively tell our story, data has to be something that is understandable, and our community can latch onto [it] and understand.”

Before using LiveStories, the agency had just a static web page that was updated weekly. It now breaks down what’s being seen in the county by flu type, age and more, while mixing in its prevention message.

The tool has helped the public and the media understand what’s happening locally, Wall says.

“Our community really enjoys data and really wants the full picture of whatever it is we’re dealing with. Especially this year when a lot of the national news was about a really bad flu season, our community was a little confused because we weren’t having that bad of a flu season,” Wall explains. Though numbers were high, they weren’t comparable to other parts of the country.

The agency can’t claim the tool helps prevent the flu, but “the conversations [Wall is] hearing are based on shared data and shared understanding,” she says.

The agency plans to use the tool with other data sets, such information around the opioid crisis, in the future.

Challenges Persist for Real-Time Flu Tracking

Going digital has its hurdles, however, including a limited historical record, Fox notes. Digital data sources have only been around for 10 to 15 years or so, which limits longitudinal studies.

Further, Fox points to one of the problems with Google Flu Trends: Its algorithms couldn’t decipher whether people Googling information on the flu were actually sick or just interested in it.

“These types of events are important because there are these anomalous events that you need to have an understanding of so your models and predictions are consistent throughout the years, even when these bizarre things happen,” he says.

Nevertheless, he expects predictions to improve as more data sources become available and such anomalous events are better understood.