More than two billion people worldwide still use Facebook for the daily exchange of personal stories, thoughts, and updates within their networks. A recent study has now tried to gather more data on the possible health problems of users by using their language and status messages: With confirmed accuracy!
The researchers reviewed the language and wording of around one million Facebook status messages. These were posted by a total of 999 users. At the same time, the medical records with demographic data (age, sex, and race) and the diagnoses of the participants were also available. The scientists divided the words found into 200 groups and used these to investigate whether certain words or word groups could be linked to specific diseases.
And indeed such connections were found: Alcohol addicts often used words like "drunk" or "bottle" or "yeah", whereas depressive people talked more about "stomach", "head" and "tears". Diabetics, on the other hand, were more likely to move in a religious environment with words like "God" and "prayer".
Interestingly, many of the clinical diagnoses examined could actually be identified by language. This worked particularly well for alcohol abuse, hypertension, and depression. For ten of the diagnoses, Facebook was even more accurate than just the demographic parameters from the databases. The researchers found the highest predictive power based on language and choice of words for pregnancy, diabetes mellitus, anxiety disorders, and psychoses as well as depression.
The study authors even have a name for the use of health-related data and language in social networks: We are currently seeing the birth of the so-called "mediom". Genomics, proteomics, metabolomics and now also mediomics? Or is everything just one step closer to "fully transparent individuals" in the end?
Merchant RM et al., Evaluating the predictability of medical conditions from social media posts. PLoS ONE 2019; 14(6): e0215476; https://doi.org/10.1371/journal.pone.0215476