The concept of mining our social media data to derive insights into our behavior is certainly not a new one, but with mental health an increasingly pressing societal concern, new research from Penn Medicine and Stony Brook University highlights how valuable Facebook data can be in providing early warning signs of a range of conditions, from anxiety to depression.

The authors believe that, with the patient’s consent of course, social media data could be monitored by their healthcare provider to look for signs of mental and physical illnesses, and prove to be as effective as monitoring for physical symptoms.

“This work is early, but our hope is that the insights gleaned from these posts could be used to better inform patients and providers about their health,” the researchers explain. “As social media posts are often about someone’s lifestyle choices and experiences or how they’re feeling, this information could provide additional information about disease management and exacerbation.”

Data mining

The researchers recruited around 1,000 volunteers who agreed to link their medical record data with their Facebook accounts so that their entire posting history could be harvested by the researchers, who then built a number of models, one of which analyzed the language used in their Facebook feed, and another two that combined data from both the medical records and Facebook.

The results were quite amazing, with Facebook data alone sufficient to allow 21 different conditions to be accurately diagnosed, with 10 of these diagnoses more accurate than using demographic information alone.

“Our digital language captures powerful aspects of our lives that are likely quite different from what is captured through traditional medical data,” the researchers say. “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.”

The work provides another telling reminder of the power of data to inform health-related decision making.  The results were strong enough using just Facebook data that you can only imagine how powerful they could be when combined with official medical data, data from mobile apps and wearable devices, potentially even genomic data.  Change is unfolding incredibly slowly in healthcare, with the promise of better use of data occurring at glacial pace, but there continues to be an ever growing number of examples of just how powerful it can be.

For now, we’ll just have to make do with another reminder of the promise that’s awaiting us.