Every day, in myriad ways, people tapping out their posts on Facebook, Twitter, Instagram and other social media may be sending out more information that even they imagine.

What if those posts, taken together, constituted a kind of trail not just to their psyches, but to their overall health? What if researchers could use this to find ways to improve health care?

That's the focus of research led by the newly created Penn Medicine Center for Digital Health.  Its goal is to study and understand the vast amounts of information on social media and how it relates to health.

Its director, Raina Merchant, an assistant professor of emergency medicine, first waded into the digital health realm when she led a crowdsourcing contest to find locations of automated external defibrillators in Philadelphia. Most people know they exist, but where's one when you need it? Using the data submitted, Merchant and her team created a mobile app that maps the lifesaving devices.

In  2012, the Robert Wood Johnson Foundation named Merchant one of the top 10 leaders in national  health care under the age of 40.  She spoke to us recently about the new project.

You use the term "social mediome." What is it, and why is it important? 

Increasingly, there are so many "omes" out there. The genome. Microbiome. Metabolome.  The focus is on the different features that tell us more about our genetic and biological makeup.   But we now have an incredible amount of social data that also tells a lot about us.  It's all the information that we post, that we share, when we check in and say we're at the grocery store or the gym, when we take photos at certain locations. If we had the ability to stitch all this information together, we would probably have a really good picture of all our different exposures and thoughts and feelings in a way that' s different from how we traditionally think about medical information.We've been really interested in trying to pull together some of this data. We call it the social mediome, as a collection of all the social data people are distributing and  sharing on a regular basis. And we're trying to better understand the connections with health.

It's early work, but we're really excited about the potential. This information in the past wasn't observable. But now we're in an era of sharing. So we're trying in a responsible and ethical way to see if there are important things we can learn about health and health care.

How are you going about it?

We have about 5,000 patients who have agreed to share both their social data and their health records.  As with all the research we do, privacy and consent are important. We've thought a lot about those things.We are only looking at things in aggregate. We aren't looking at an individual's daily tweets. Rather, we're looking at a group of individuals who have a certain health condition. Are there similarities in the types of things they post or share?

Obesity is an example. We would have thought that patients who were obese might post more about diet, exercise, food, recipes.  What we found is that it was a little more subtle.  The language was more about being sedentary.  The words people used were about "sitting" or "being still" or "being planted."

Some things we find are more along the lines of what we might expect. People with depression were more likely to post about feeling sad or blue or crying or being upset.  People who did not have a diagnosis of depression were less likely to use those words.

It's still very exploratory. It's about having a large sample of data to find patterns that will tell us more about health.

Why not just ask the people themselves how they're feeling?

There are different ways in which we collect information and understand disease.  The traditional one is that we ask people how they're feeling. And that's still important. But, potentially, the social data gives us these earlier clues.  Someone may not realize that they're depressed, but their language or their actions might reveal that.  One study shows that people with a diagnosis of depression use different image filters on Instagram.  They use more grays and darker shades of color.

Another area that we think might be interesting – and this is still very early – is with people who have cognitive decline.  Potentially, if you had a writing sample from them on a regular basis, you might see that the language  or the complexity of sentences might change over time in a way you might not be aware of otherwise.

You're also looking at Yelp posts that rate hospital experiences. Tell us about that. 

We're trying to better understand more about the patient experience from information posted online.  What do patients care about most when they post about a hospital?  So far, we have looked at about 45,000 reviews of hospitals online. We have identified themes that are important to patients. And not just patients, but also their family members and loved ones.  Often, the family member is the one who is there day and night, talking to the doctors and nurses and really experiencing a hospital in a way that might be different from that of a patient who is not feeling well.

In traditional surveys, patients are asked more broadly about the overall hospital experience, but not about the care that they might receive in specific areas of the hospital.  In a Yelp review, people might talk about the care they receive in the emergency department or the labor and delivery floor.

A lot of their reviews are really positive. It's not that people just go onto Yelp to write about a one-star experience. I think there's also a lot to be learned from what was a really amazing experience, as well as  what was pretty awful.

Ultimately, where is this headed?

As the world evolves, health care is changing rapidly. Our hope is that the digital platforms and tools now at our disposal offer great opportunity for thinking about health care differently — how we connect with patients and how we receive and share information. If we can develop approaches for testing and vetting and studying new tools as they arise, we can be on the cutting edge. In the future, we hope to have more precise tools that allow us to understand more about the care of the individual patient. To get there, we need to find a signal among a large group of patients.

With the social mediome, the more language and digital footprints that we collect, the more samples we have, the more nuanced findings we can understand between digital footprints and health.