Twitter Review Reveals Clues as to Who, When, and What Users Are Tweeting About Cardiovascular Disease Online

For many clinical researchers, social media is a gold mine of information just waiting to be tapped by science and, according to a new study, Twitter might be the place to start for understanding cardiovascular disease trends.

The ubiquitous website that allows users to send 140-character Tweets out to followers around the world has the potential to answer many questions related to global health, lead study author Lauren Sinnenberg (University of Pennsylvania, Philadelphia), told TCTMD. “We wanted to know if there is a signal in all of that data,” she said.

Her team reviewed more than half a million Tweets sent out between 2009 and 2015 that were associated with cardiovascular disease. Among them, terms like diabetes (n = 239,989) and myocardial infarction (n = 269,907) were used abundantly, while terms like cardiac arrest (n = 12,238) and heart failure (n = 9,414) were seen less often. Additionally, people who tweeted about cardiovascular disease were generally older (mean age 28.7 vs 25.4 years; P < 0.01) and more likely to be female (52.7% vs 51.2%; P < 0.01) than the overall population of Twitter users.

Within a sample of 2,500 tweets that were manually analyzed for content, common themes centered around risk factors (41.9%), awareness (23.4%), and treatment and management of disease (21.6%). Among the 9.9% of tweets that mentioned outcomes, 78.1% mentioned death. Interestingly, metaphor was used in 44.2% of tweets related to cardiovascular disease, emotional language in 39%, and first-person accounts in 34.9%.

The findings were published online today ahead of print in JAMA: Cardiology.

Misspellings, Slang Muddy the Results

In an accompanying editor’s note, Mintu Turakhia, MD, and Robert Harrington, MD (both  Stanford University School of Medicine, CA), admit that this study is unlike others that have been published in the journal before, saying that “digital health, broadly defined, is in its infancy.

“We accepted it because it highlights the potential for using these emerging data sources such as Twitter for cardiovascular research, in this case to evaluate public communication about cardiovascular medicine in a manner not previously possible on such a scale,” they write.

One limitation to the study, Sinnenberg acknowledged to TCTMD is its inability to capture all slang and misspellings for the various terms related to cardiovascular disease the researchers searched for. “We used many, many search terms, but the list is exhaustive,” she said.

Because of this, C. Michael Gibson, MD, (Beth Israel Deaconess Medical Center, Boston, MA), who was not involved in the study, told TCTMD that “it becomes really hard to turn this into a quantitative research tool as opposed to a qualitative kind of descriptive research tool.” Much work remains to be done on “natural language processing,” as, for example, there are more than 700 different ways people refer to aspirin, he added. “It would really require a vast undertaking to really get all that right in terms of identifying things.”

While encouraged that people are tapping into the data, Gibson said, “I think this is just so nascent that it's so far from being useful at this point.”

Next steps for research would be building terminology libraries to account for the various ways people talk about health and medicine clinically and casually, he said. It might also be helpful to create “tools where everyone who is a cardiologist registers as a cardiologist and agrees to allow their Twitter conversation to be tracked and used for these kinds of purposes,” Gibson suggested.

The problem of overlap also needs to be solved, he said, giving the example that MI could stand for both myocardial infarction and mitral insufficiency. “Context of search terms becomes very important,” Gibson observed.

Potential for Wider Implications

But once all the tools are created and researchers have a better handle on how to interpret data from social media, Sinnenberg said, questions related to patient comprehension will be easier to answer. “Folks talk about their disease to their physicians in the doctor’s office and that's a really clinical, sterile setting where people use one specific type of language,” she observed. “But Twitter really allows us to get into how people talk with their families and friends in a candid environment so we can really understand a lot about how folks who have these diseases understand their diseases by tapping into this.”

In the future, this information “can help physicians understand how to communicate with their patients about their diseases,” she said.

It is also interesting to note “how salient these diseases are to patients—that they are talking about these diseases with their friends and family and that they're sharing the knowledge they get when they are in the physician's office and that they are spreading it around,” she said. “It highlights the importance of those moments of communication, and the patient-provider relationship to understand how that message is propagated.”

Additionally, Gibson said that information gleaned from social media would help on a larger public-health scale in making it easier to detect disease outbreaks worldwide. “This would be a great tool to be an early alarm for the migration of different flu variants,” he said. “On drug and device safety, it could be a spectacular tool to track device failures [and] drug reactions, things like myalgias with statins as an example, to really get a better sense of the true prevalence of these kind of adverse events.”

Other potential uses for social media in medical research, Turakhia and Harrington suggest, could relate to “monitoring of prescription medication abuse, recruitment and conduct of clinical trials, and as an intervention to improve caregiver and peer support for, education in, and management of chronic diseases.”

But for now, while this research is “fun and exciting,” Gibson said, “it has a long way to go.”





Related Stories:



  • Sinnenberg L, DiSilvestro CL, Mancheno C, et al. Twitter as a Potential Data Source for Cardiovascular Disease Research. JAMA Cardiol. 2016;Epub ahead of print.

  • Turakhia MP, Harrington RA. Twitter and cardiovascular disease: useful chirps or noisy chatter? JAMA Cardiol. 2016;Epub ahead of print.

  • Sinnenberg, Gibson, Turakhia, and Harrington report no relevant conflicts of interest.