Facebook posts could be a window into more than just what you did last weekend. It turns out the language used in a Facebook post can predict if a user has a certain medication condition, according to a new study published in PLOS ONE.
Academics from the University of Pennsylvania School of Medicine set out to research if Facebook posts could predict conditions that were already recorded in a patient’s medical record.
“In what we believe to be the first report linking [EHR] data with social media data from consenting patients, we identified that patients’ Facebook status updates can predict many health conditions, suggesting opportunities to use social media data to determine disease onset or exacerbation and to conduct social media-based health interventions,” authors of the study wrote.
Researchers found that Facebook language was able to predict 21 medical conditions “beyond chance.” Of those, 18 conditions were best predicted using combination of demographic data, as well as Facebook post language. Scientists also discovered that 10 of those conditions were better predicted by Facebook language than by standard demographics including age, race and gender.
Facebook language was particularly accurate at predicting which participants were pregnant or had diabetes, anxiety, psychoses or depression.
“We found that the language people use in Facebook is predictive of their health conditions reported in an [ERH], often more so than typically available demographic data,” authors wrote.
HOW IT WAS DONE
The team of researchers used a total of 949,530 Facebook status updates across 999 participants. In order for a participant to be included they had to have at least 500 words of status updates over the course of a two-year window.
The participant’s language was put into a natural language processing system. Researchers built three predictive models: one used only facebook language, another used only demographic information, and a third that combined Facebook language and demographic data.
WHAT'S THE HISTORY
Medical professionals have turned to social media before to predict various health-related trends.
“Social media content has been shown to contain valuable health signals, though mostly at the population level. For example, Twitter has been used to surveil disease outbreaks, predict heart disease mortality rates and to monitor public sentiment about health insurance,” authors of the study wrote.
ON THE RECORD
"Our digital language captures powerful aspects of our lives that are likely quite different from what is captured through traditional medical data," Andrew Schwartz, an author of the study, said in a statement. "Many studies have now shown a link between language patterns and specific disease, such as language predictive of depression or language that gives insights into whether someone is living with cancer. However, by looking across many medical conditions, we get a view of how conditions relate to each other, which can enable new applications of AI for medicine."