In 2008, Google researchers pioneered the idea of using specific keyword searches to track the location of seasonal flu outbreaks. Working in cooperation with officials at the Centers for Disease Control, the company ultimately released Google Flu Trends in 2010, a tool they hoped would complement traditional CDC data in predicting flu outbreaks.
After its initial successes in predicting the scope of the H1N1 influenza outbreak, Google Flu Trends hit a roadblock, grossly overestimating the number of flu cases in the 2012-2013 flu season. This misstep may have demonstrated the limitations of search query data analysis in predicting disease.
Enter Social Media Research
Image via Flickr by Kooroshication
Graham Dodge, entrepreneur and founder of Sickweather, believes social media chatter is a better indicator for predicting disease outbreaks. Social media sites like Facebook and Twitter provide real-time contextualized data that can be mined and analyzed to accurately track and predict disease. His company’s new consumer product is doing exactly that.
Social media, according to Dodge, “provides more context … for natural language processing to better qualify what the person means.” He explained there is more predictive data in a tweet saying “I have the flu,” for example, than in a Google search for the word “flu.” His tracking product analyzes social media data for mention of 24 symptoms of illness and then plots them by geographic location to identify “sick zones.”
Impact of Social Media Research on Public Health
The results of social media research efforts caught the attention of Dr. Nicole Lurie, the Department of Health and Human Services’ undersecretary for preparedness and response. DHS announced a contest to develop the best public health app to predict disease outbreak based on social media feeds, especially Twitter.
In September 2012, the $21,000 cash prize was awarded to two nursing informatics specialists, Charles Boicey and Brian Norris, for their web application “MappyHealth.” The application focuses strictly on Twitter data and incorporates robust geolocation information so public health officials can focus on their local and regional populations as well as tracking global trends.
The MappyHealth developers went on to found Social Health Insights, a company that uses social media analysis to create interactive maps that pinpoint which diseases are being discussed in which U.S. cities.
Integrating Social Media Research with Traditional Data
The CDC is already incorporating search query data and social media data with its traditional data tracking emergency room visits and cough and cold medication purchases. But Sickweather’s Dodge believes integrating a broader range of data will be more accurate in forecasting disease outbreak. He cites the period around the Indianapolis Super Bowl, during which time Twitter mentions of flu and flu symptoms skyrocketed in the Indianapolis area.
Dodge’s product combines major event information, weather and travel patterns, and environmental data with social media analysis to more rapidly and accurately identify sick zones and outbreaks.
Integrating social media research into healthcare informatics is a concept well worth developing. With over 500 million tweets sent each day alone, social media offers a treasure trove of data that can be mined to help protect public health.