skip to main content
Global Search Configuration

As the use of social media continues to rise and become more ubiquitous, researchers are determining how user data can be used in health-care screening. Medtech Insight explores how social media could be used to predict and manage patient health in the future, including research on its use to improve diabetes management and its potential in mental health assessments.







Today, more and more of us are embracing social media to communicate at home or at work. Whether it's Facebook, Twitter, Instagram, or some other platform – the power of social networking is ever-present in our lives.



So much so that according to a report by eMarketer, by 2020 the number of worldwide social media users is expected to reach some 2.95 billion, around a third of the global population. With social media becoming the dominant form of communication, health-care stakeholders are turning to this information being generated as a means of gaining insights into health and management of disease.



In a recent study, Michelle Litchman, a specialist nurse practitioner and assistant professor at the University of Utah College of Nursing, Salt Lake City, used the social media platform Instagram to search images of people using continuous glucose monitors (CGMs) made by Dexcom to see where the monitors were being worn on the body.



At present, US FDA has approved Dexcom monitors for use on the abdomen. However, Litchman was finding that patients coming into her practice were wearing their CGMs on other parts of their anatomy. "When I asked patients about it, they said they felt that they were getting really good, accurate readings and had seen others using these alternative [anatomical] sites online," explained Litchman.



She decided to investigate this further and together with a team of colleagues, Litchman did a manual search on the Instagram website for photos of patients wearing Dexcom CGMs. Searching under the hashtag 'Dexcom', they found 353 public posts featuring people wearing Dexcom monitors, which is one of the leading devices of its kind on the market. The team coded the data based on which body part the CGM was being worn and also coded the comments on each post. Litchman observed that users were providing "peer-to-peer education" on social media, often teaching other users how the CGM might work for them when worn on different sites or explaining if it was comfortable in that position.



The team determined that two out of three Dexcom CGM users wore the device on an 'unapproved' part of the body but had good results. In addition to the Instagram study, Litchman and her team submitted a manuscript on the use of CGM and how patients share their data with care partners, such as a parent or spouse. The study used information acquired from online diabetes blogs. "We wanted to get the view from both sides, so not just the person with diabetes but also the care partner," Litchman told Medtech Insight. "We did a systematic appraisal and examined 39 different blogs that were in the diabetes online community, looking at 206 comments on the blogs in the comment section." The research demonstrated that mining personal blogs, provided useful insights into patient's diabetes management.



With Litchman's team currently conducting all research manually, the next step will be to digitize and automate these searches. "One of the challenges of a digital search on Instagram is ensuring that different shadows within photographs don't stop an analyzer from detecting a device. Of the photos we found, sometimes the CGM was in a shadow or at a slightly hidden angle. This could be a challenge, going digital, but I definitely think it can be done," she said. "It would also be interesting to look at other hashtags and potentially other social media sites, because we know that Instagram has a much younger audience so we could explore Facebook, Twitter and even Tumblr that attract different types of people."



Mental Health Promise – But Data Privacy A Snag


Other researchers are employing artificial intelligence and deep learning techniques as a more efficient way to sort through large clusters of user-generated information. Scientists from Harvard University and Vermont University recently published a study in the in the journal EPJ Data Science, showing how machine learning was able to identify markers of depression. Using Instagram data from 166 individuals, the researchers analyzed statistical features computationally extracted from 43,950 participant Instagram photos, using color analysis, metadata components, and algorithmic face detection.



Utilizing machine learning, the scientists were able to identify distinctions in color choices, image qualities, and filters and enhancements among participants in both categories. These profile patterns were analyzed to create a model confirming or predicting depression.



Depressed and undiagnosed depressive individuals tended to post darker images with blue or gray tints, blurred images, and single-face photos more often than their healthy counterparts, who posted brighter or lighter images with more aesthetic enhancements, and multiple faces in a photo. Additional findings revealed that depressed participants generally received more likes on their photos and posted more often.



Andrew Reece, co-author of the study from Harvard University cautioned that despite results suggesting the promise of using these social media photos for early screening and detection of mental illness, the science was at too early of a stage to understand the long-term potential for diagnostic use.



"I think there are a few important caveats that need to be considered when we talk about how predictive mental health screening might be used," said Reece. "First of all, there's a concern about exactly who are the people we've got to participate in the study and are they really representative of the broader population - or is there something specific about them that we've just learned through this study?"



To participate in the study, a select group of individuals had to fulfil set criteria. Participants were crowdsourced using Amazon’s Mechanical Turk (MTurk) crowdwork platform and had to complete separate surveys for depressed and healthy individuals. In the depressed survey, participants were invited to complete a survey that involved passing a series of inclusion criteria, responding to a standardized clinical depression survey, answering questions related to demographics and history of depression, and sharing social media history.



"We had far more people drop out of our study at the point when we asked them to share their social media profiles compared to when we asked them to share their personal mental health history," said Reece. "43% of all our participants walked away when we asked them to share their Instagram photos. So to me, that's a strong signal that data privacy is an extremely relevant issue and there isn't a whole lot of trust between the general public in how researchers or companies may use their data."



To overcome this obstacle, researchers must establish more trust from the public in how personal data will be used in studies. "In the future if there was going to be some kind of set-up where physicians were able to access patients' social media data to make assessments, this would definitely have to be voluntary and the patient would need to have total control over how their data is used," said Reece.



In addition to the Instagram findings, Reece previously conducted a study using Twitter to forecast and trace the onset of mental illness. Twitter data and details of depression history were collected from 204 individuals, with the team extracting predictive features measuring linguistic style, and context and then built models using these features with supervised learning algorithms. The resulting models successfully discriminated between depressed and healthy content, and compared favorably to general practitioners' average success rates in diagnosing depression.



"In the graphs from the study, 6-9 months before people received their official diagnosis for depression there's a line for healthy and depressed people. The lines start off at the same point and then suddenly, just based on the language used in the Tweets by depressed people online the probability of depression based on what you're saying online rises and peaks just when depressed people get their first diagnosis," Reece told Medtech Insight.



"Then, about 8-12 weeks after that diagnosis, we begin to see the probability and the line go down for the depressed people and that's right around the time when therapies start having an effect. It's fascinating to detect a signal of if there's a problem and watch it progress over time. With this technology, we can see it well before an individual goes and gets an official diagnosis."



With the advancement of this technology in health care, governments would have to play a key role in regulation. "Even public data can unintentionally contain private information if you have the right analytical tools," says Andrew Reece, Department of Psychology, Harvard University. "So, I think the role of science, as always, is to be rigorous and careful in discovery of facts and new ways of learning about the world. Then the role of government is really to keep in check the way that progress grows and provide ethical guidelines in place governing privacy. Scientists and government officials need to work hand in hand."



However, Reece said with the advancement of this technology in health care, governments would have to play a key role in regulation. "Even public data can unintentionally contain private information if you have the right analytical tools. So, I think the role of science, as always, is to be rigorous and careful in discovery of facts and new ways of learning about the world. Then the role of government is really to keep in check the way that progress grows and provide ethical guidelines in place governing privacy. Scientists and government officials need to work hand in hand."



Twitter For Predicting Flu Spread


Twitter has been used previously to predict the spread of flu. In 2013, the US Centers for Disease Control and Prevention launched the "Predict the Influenza Season Challenge," a competition that encouraged researchers to use social media to post their prediction on how the flu will spread. Researchers at Northeastern University, Johns Hopkins University and the University of Rochester found real-time information like tweets posted on Twitter to be a useful source of public health information and that location-based tweet could be used to track outbreaks and actually predict where flu will spread.



This information could be invaluable for public health departments, providing them advanced warnings and time to plan with additional doctors or hospital beds. With GPS information embedded in tweets, the location information gathered from Twitter is vital for pinpointing outbreaks.



But is the hype around these applications outpacing the reality or are we set to witness a turning point in predictive health-care screening? "In the case of depression, an accurate diagnosis is very difficult to get, especially when there are many other things that can be going on," said Reece. "If there's some kind of technological solution that people are willing to participate in and share all of this data that they've created, with possibility for it to feedback and help them one day, that's really where we need to show a proof-of-concept."



For Litchman, she believes the value of social media will be fully realized by researchers and health-care community once more studies are conducted. "There's a lack of research validating the use of social media in a more randomized, controlled setting, which is what we need to be doing," she said.



"One of the things we found in practice, was diabetes patients who were engaged with the diabetes online community via social media, had a better A1C [blood glucose level] test, so now our team is looking at a small trial in which people will be randomized in either being part of a diabetes online community or control group to see the effect. I think studies like that are going to help provide more data for health-care providers in determining whether social media provides value and what value it can provide."



Ultimately, for patients and health-care providers, social media will always be a useful tool for providing patient community and support. Litchman said: "We already see social media being used day to day for the support, tips and tricks for people with diabetes. On Twitter, there's a tweet chat every Wednesday night and there's a lot of people that engage with that, people asking questions and offering advice and support. People raise each other up through social media by getting emotional support and a community.



"If you're part of a community where other people are similar to you then it's positive. At the end of the day, collective wisdom of having several people who are similar to you is always valuable in getting the information that is helpful to you."

Read also


Next steps

Whether you’re a small biotech start-up, research firm, generic manufacturer or a global pharmaceutical giant, you need focused, independent insight and opinion on market developments.

Our team is ready to hear from you for a particular request or area of interest. Please do not hesitate to reach out and discuss.

Contact us for product technical and account support.

  • US Toll-Free   : +1 888 670 8900 
  • US Toll             : +1 212-600-3520
  • UK & Europe : +44 (0) 208 052 0700

Have an immediate and specific information need?

Browse and buy from 1000s of analysis and research reports now: