I had the honor of writing my first guest post on “the first blog in medicine” – KevinMD.com – last week. That post was written in advance of my announcement at the Mayo Clinic’s Scientific Session about the debut of MDigitalLife.com – the first known database of twitter-using physicians to matched to an “official” data source – the NPI (National Provider Identifier) database. That post is reprinted in its entirety below.
It’ll be no surprise to the readers of this blog that physicians’ use of Twitter and other social media has been exploding over the last couple of years. But it may surprise you to know how hard it is to really analyze that data.
Last year Dr. Katherine Chretien of the VA Medical Center in Washington, DC, published an eye-opening study in a JAMA letter. Until that point, all we really had were anecdotes and survey responses – certainly not the same as analyzing what physicians were actually doing and saying on Twitter.
By leveraging a strong research team, Dr. Chretien was able to narrow a list of 523 potential author candidates to a final group of 260. The team then looked at the last 20 tweets from each author – a total of 5,156 were included in the study. This study was and is a big deal. It answered a lot of questions – but perhaps the most important one was whether physicians, when using social media, would be more prone to share confidential patient information, or to share anything unprofessional in nature (e.g., used profanity or made discriminatory statements).
I know that I breathed a sigh of relief when the team found that only 3% of tweets contained any unprofessional material. Don’t get me wrong – any amount would be too much. But as a physician friend of mine likes to say, “we’re a lot worse at the water cooler and in the elevator than we are online.” The point is that by publishing a credible study in a credible outlet, Twitter for physicians had become suddenly more acceptable.
My colleagues and I have been following thousands of doctors on Twitter for the last few years … and have gotten more and more interested in finding out what we could learn from them if we were to try to automate some of the very difficult process of sourcing and analyzing their Twitter posts at a much larger scale.
To that end, we have created a database of nearly 1,400 U.S. doctors (to date) on Twitter and – this is critical — linked them to NPI (National Provider Identifier). To our knowledge, this is the only database linking doctors using the Twitter platform to an “official” data set. We’ve captured over 400,000 tweets from those doctors over a 5-month period in 2012, allowing an unprecedented ability to scan for trending topics and relationships, broken out by specialty, gender and location.
- The specialty mix is a near-perfect match with the overall US physician population; no specialty varied by more than 3% from between our data set and the country’s doctors as a whole.
- Our geographic mix was also very close; 49 states indexed within 4% of the overall physician population in that state. Only California strayed outside that number; Californian physicians over-index by 7% in terms of their Twitter usage.
Having been satisfied that we had assembled a reasonable data set, we’ve spend the last two weeks poring through that data looking for insights and patterns. Here are a few of our top-line findings:
- These are active users. They tweet over 2x per day on average.
- Twitter is a part of their work-day. More than 50% of tweets are sent between 9am and 5pm (in the physician’s local time zone)
- They have an audience. 2/3 have at least 150 followers (the median is 306)
- They connect to each other. More than 1/3 of the doctors are followed by at least 20 other doctors in the database.
- The most-followed physician by those in the database was none other than Kevin Pho; followed by nearly half of the doctors studied.
What’s more interesting, of course, is the substance of their tweets. We examined a sample of tweets related to 3 therapeutic areas: diabetes, breast cancer and prostate cancer. Perhaps not surprisingly, 83% of specialties mentioned diabetes in their tweets … though it’s interesting to note that pediatricians discussed it among the most frequently (behind only internal medicine and family medicine doctors). I suppose that’s a sign of the times – childhood obesity and the earlier onset of diabetes is very much on the mind of the physicians who care for our kids. It was also interesting to note that prostate cancer was as widely discussed as it was … 70% of specialties mentioned it in their tweets. Much of the volume there was driven by this summer’s controversy about the validity of the PSA test as a means of lowering morbidity; In fact, 43% of the tweets about prostate cancer between May and September were related to this specific issue.
Although we’re only six weeks into what will almost certainly be a years-long process, it’s been fascinating to discover – in a scientific and quantitative manner – how innovative physicians are leveraging social media tools like Twitter to improve the health of their patients, to grow their practices, to connect with their peers, and to create a platform for influencing the future of healthcare.
We’re already looking to work with some of the most prestigious healthcare institutions in the world, including the Mayo Clinic, to glean deeper insights from this data. But as I go into my first presentation of this material, I’d love to know: What are your biggest questions? What are the thorny problems you’re trying to solve that would benefit from a deeper look at the physician’s perspective? It’s my hope that we’ll be able to solve some of them together.