Guest post by Dr. Oliver Boegler
A Numerical Analysis of a #BCSM Tweetchat
At ASCO this year I was honored to participate in an educational session entitled: “Internet, Social Media, Privacy Regulations, and Clinical Trials”. While it was a little daunting appearing alongside Michael Thompson, MD, PhD (@mtmdphd) and Robert Miller, MD (@rsm2800), both very well known experts in the area of social media and clinical trials, I had a great time describing my experiences on social media from a patient’s perspective. An important part of that for me is #bcsm so I naturally spoke about it. Here I want to share what I said about #bcsm with you.
Long before ASCO, I was invited to participate as an invited guest one #bcsm Monday night to discuss male breast cancer. The session was scheduled several months in advance, and took place October 7, 2013. At the time I was first invited, I wondered to myself how interesting a session on the male disease might prove to the #bcsm community, and so set out to gather data that would allow me to look at this. In short, I collected all the tweets with the #bcsm tag for several months either side of the male breast cancer session, and analyzed them. Here is what I found.
Of course, before diving into the analysis I introduced #bcsm to the audience at ASCO, and its founders. Dr. Attai (@drattai) was in the audience also!
Then I briefly described my methodology for gathering the data. I used a webservice called If This Then That (IFTTT) where you can make simple scripts that combine webservices. In this case every time a tweet with #bcsm posted, it would be written to a google spreadsheet in my google drive. I collected 24,000 tweets from July to February.
Then I combined the spreadsheets that collected over the months, fixed some errors in the data and loaded it into my favorite data visualization software, Tableau. The first thing I looked at was the pulse of the community. Given that the tweetchat is on Monday night, I wanted to see when during the day and during the week people posted with #bcsm. What you can see is that most posts are during US daylight hours, with an identifiable peak around 8pm (times are Central because I live in Houston, and google knows that).
You can also see that Monday is the day when most tweets are posted, again not surprising. What was surprising to me was that there was so much activity throughout the week, with only about 20% of all posts occurring on Mondays. Also interesting is that although October is clearly among the busier months, its not the most active, at least in this particular slice of time.
Next I asked how participation is distributed amongst the tweeters who use the tag. What stands out is that there is a very long tail of people who participate, but post infrequently. Happily our three founders were amongst the most active tweeters – phew. Even the most active members of #bcsm contributed only around 2% of the tweets. This suggests what we know from our experience – the community is wide open for participation.
I wondered next what the role of re-tweets is in the pulse of #bcsm. Interestingly, 60% of tweets had the keyword RT in them, and the striking peaks in total activity appear to be due to the volume of RT activity. You can see this by examining the line graph below, where the blue line, which marks original tweets, oscillates within a well-defined band over the period of analysis. However, the RT frequency, shown in orange, shows some unusually high peaks and these mark the most active sessions. One conclusion is that when things resonate with the participants, motivating people to RT, the volume spikes. If correct, this suggests that RT volume may be a reasonable measure for the level of interest in a subject. An alternative explanation is that the request by our moderators to keep RTs to a minimum, which is usually made at the start of the tweet chat, is heeded to a greater or lesser extent for reasons unrelated to the level of interest in the subject. Perhaps when Twitter is slow, people RT less to reduce load on the servers, and so we may instead be looking at a surrogate of Twitter’s technical performance. As a good scientist I conclude that this needs more study. (A grant for twitterology anyone?)
Another thing you can see in the analysis is that different tweeters use RT in different proportions. Two of our moderators, for example, both RT and write original tweets.
Then I focused (finally) on the session on male breast cancer. I tried (unsuccessfully) to convey the pace of a tweetchat by showing a movie of the session transcript whizzing by (movie didn’t play). You all know what its like to participate in #bcsm so I don’t need to belabor this.
One a way the session can be described with is the sentiment that Social is Education made by many, and recently in a tweet from Bernadette Keefe – I feel we talked about a series of issues related to the male disease that not everyone was familiar with.
So do we tweet about the male disease generally, or is it more focused on when it is a formal topic? An analysis of tweets that contain either the word “male” or the word “men” shows a fairly flat line with a noticeable peak on October 7th (the small orange trace below the big blue line, which shows total #bcsm volume, as above), suggesting that it does come up occasionally, but of course most when it’s the focus. Overall less than 2% of the tweets with #bcsm contain these key words, which is pretty reasonable given that the incidence of the male disease is about 1% of breast cancer patients. I also identified the tweeters who use these words the most.
Interestingly the day of the male breast cancer tweet chat was one of the days that showed a definite spike in RT activity, and an overall strong volume (see above). This raises the possibility that the #bcsm community found the topic interesting, perhaps because it is not a common focus of conversation. There are alternative explanations, as mentioned above, but I favor the interpretation that the discussion generated significant interest and sharing of information and so was effective in our goals of raising awareness and educating.
As a control for the analysis of “male or men”, I next looked at the same question for the keyword “clinical trial”, and a relatively similar pattern emerges. A noticeable spike was seen at the time that this was the topic of the tweetchat. Gratifyingly Dr. Thompson, very active in this area and a regular guest at #bcsm was third most frequent in this analysis. About 1% of tweets with #bcsm also contain “clinical trial” and the topic does come up every once in a while.
In closing, I conclude that the #bcsm tweetchat on male breast cancer showed relatively high tweet volumes, when compared to the time period – it was only beaten by two other peaks in the graph. This high volume was largely due to RT activity. Together this could be interpreted to mean that it was a subject that was interesting to the community, and that the information discussed was found to be worthy of sharing by the participants. As such it was most likely a successful attempt to raise awareness and educate.
I thank the #bcsm community for their willingness to discuss this topic and for the opportunity for me to share this blog post.
Dr. Oliver Boegler is the Senior Vice President, Academic Affairs, Division of Academic Affairs, Sr VP Office, The University of Texas MD Anderson Cancer Center, Houston, TX; Professor, Department of Neurosurgery, Division of Surgery, The University of Texas MD Anderson Cancer Center, Houston, TX; and Associate Dean, The University of Texas Graduate School of Biomedical Sciences, Houston, TX. He is also a male breast cancer survivor, and blogs about male breast cancer at Entering a World of Pink.