Facebook Posts Could Indicate Medical Conditions Better Than Demographics: Study
New research has found that our social media habits could be an indication of medical conditions.
A report published in the PLOS One peer-reviewed scientific journal has found a link between the words people use on social media and the probability of suffering a certain type of medical condition.
The research found that language people use on Facebook is "predictive" of health conditions reported in their medical records, often "more so" than typically available demographic data.
"Patterns of language can be associated with diagnoses to reveal similarities and difference between diagnoses," the study said.
The research, conducted by the University of Pennsylvania School of Medicine, analysed about 20 million words used by nearly 1000 consenting patients in their social media posts. The medical records of the patients were also examined and cross-referenced with the posts.
Alcohol abuse was most prevalent in people who made posts containing the words 'drink', 'drunk' or 'bottle'.
The most predominant marker for drug abuse was posts that contained "hostility", the study said, with words such as 'dumb', 'bitches', 'bullshit' and 'people'.
Words that were markers for depression were 'stomach', 'head' and 'hurt', suggesting somatisation -- the manifestation of psychological distress by the presentation of physical symptoms.
While most medical conditions had obvious links with the words used in social media posts, one outlier researchers found was a link between diabetes and the use of religious words or topics.
The study found that people who use words such as 'God', 'family' and 'pray' in their posts were more likely to be diabetic, suggesting disease risk could be associated with socio-environmental variables.
The top 25 percent of patients mentioning religious words were 15 times more likely to be diabetic than the bottom 25 percent of people using those topic words.
"This does not mean that everyone mentioning these topics has the condition, but just that those mentioning it are more likely to have it," the study said.
The study found that the words do not represent the "causal mechanisms", but rather the findings are "correlational".
"However, in revealing what people think, feel, and do, social media patterns capture emotional, cognitive, behavioural and environmental markers that have substantial predictive validity and are otherwise fairly elusive to researchers and clinicians," it said.