Healthier Living Through Twitter

Have you ever Instagrammed your fancy new dinner recipe? Or maybe tweeted about how that burrito isn’t sitting too well? Those social media posts are more than just personal anecdotes – to the University of Cincinnati's Michael Widener they’re data points, and taken together they can offer insights into geographic patterns of public health.

Widener and his co-author, Professor Wenwen Li of Arizona State University, conducted a spatial analysis of national twitter data in an attempt to use the social media platform as a form of public health surveillance. But unlike previous studies that used social media to monitor things like the spread of infectious disease, Widener focused on mapping dietary trends.

Much of Professor Widener’s research has focused on the phenomenon of ‘food deserts,’ which the

USDA defines

as urban neighborhoods and rural towns without ready access to fresh, healthy, and affordable food. By cross-referencing the location of food-oriented tweets with geographic data on food deserts, Widener was able to explore the relationship between dietary habits and proximity to fresh, healthy food supplies.

But Widener’s study goes a step further, examining tweets not just for what people are eating, but how they feel about it. This ‘sentiment analysis’ takes the study beyond the pure geographic analysis of where less healthy dietary habits are prevalent, giving insight into how people view their own diet.

The study examined the content of roughly half-a-million tweets containing food-related terms. After filtering out erroneous posts – for instance, tweets containing the word “apple” that were about iPhones rather than eating – the remaining 200,000 or so tweets were mapped using geographic information systems and divided into categories of healthy and unhealthy, and positive or negative sentiment.

The results showed some relation between proximity to food deserts and less healthy eating choices. It also revealed that those in food deserts were less likely to espouse positive sentiment about healthy foods.

Widener is quick to point out that this is more of a proof-of-concept study than an in-depth analysis, and that there is not yet enough evidence to claim a direct correlation between geography and dietary habits. He also noted that Twitter alone is not necessarily a sufficient tool for this kind of study, since the population of Twitter skews younger and more urban than the population as a whole.

Still, in the authors’ words, “These findings substantiate the methods used by the USDA to identify regions that are at risk of having low access to healthy foods.” Widener hopes that this study will lay the groundwork for more in-depth research into utilizing the power of social media to improve public health policy in the future. The findings will be published in this month’s Journal of Applied Geography. An advance copy is currently

available online

.