MENU

A New Algorithm Uses Your Facebook Profile, Posts To Predict Your Medical Conditions

Close

What people write on social media and how they say it can tell the world so much about them, particularly about their health.

New research from the University of Pennsylvania School of Medicine and Stony Brook University reveal that the language used on Facebook posts could be indicators of existing medical conditions. With patient consent, they could be monitored the same way as 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," explained lead author Raina Merchant, MD, MS, the director of Penn Medicine's Center for Digital Health. "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."

Predicting Health Issues Through Facebook

In a study published in the journal PLOS One, researchers analyzed the entire Facebook history of nearly 1,000 patients who agreed to for their profiles to be linked to the data from their electronic medical record. The researchers wanted to see whether data from Facebook alone could indicate 21 specific physical and mental health conditions.

Three models were built to see their predictive power on the patients' medical conditions. One model only analyzed language of Facebook posts, another used demographic data of the patients such as age and sex, and the last model used a combination of the two datasets.

Findings showed that all 21 medical conditions that the researchers tested could be predicted from the data pulled from Facebook alone. Ten out of the 21 conditions were even better predicted from the Facebook data compared to demographics.

A number of the Facebook indicators that were more predictive than demographic data are simply intuitive, such as the words "drink" and "bottle" predicting alcohol abuse. The team pointed out that other indicators weren't as straight-forward. For instance, people who most often used religious language such as "God" and "pray" were found to be 15 times more likely to have diabetes than the people who used religious words the least. Hostile language are also found to be indicators of drug abuse and psychoses.

Potential Future Applications

Senior study author Andrew Schwartz, PhD of Stony Brook University, who is a visiting assistant professor at Penn, explained that people's digital language captures certain aspects of our lives differently from what is usually found in traditional medical information.

Previous research showed that Facebook posts could predict depression in patients earlier than a clinical diagnosis. The new study builds on this research, suggesting that there's potential for the development of an optional system, in which a patient's social media posts are analyzed for additional, potentially valuable information.

Merchant explained that it's not certain how widespread this system would be, but it could provide valuable information on patients who use social media frequently.

"One challenge with this is that there is so much data and we, as providers, aren't trained to interpret it ourselves — or make clinical decisions based on it," Merchant said. "To address this, we will explore how to condense and summarize social media data."

ⓒ 2018 TECHTIMES.com All rights reserved. Do not reproduce without permission.
Real Time Analytics