Skip to content

Crowdsourcing healthcare social media analytics with #hcsmca

September 7, 2012

On August 29, Symplur co-founders Howard Luks, MD (@hjluks), Tom Lee (@tmlfox) and Audun Utengen (@audvin) joined the #hcsmca chat to ask about learning from health care social media analytics. Tom shares their take-aways. 

Guest post by Tom Lee

The #hcsmca community and the healthcare hashtag project both appeared on the scene nearly two years ago in late 2010. Since that time we’ve all seen an explosion in the use of Twitter in healthcare, and with that growth the conversations have become increasingly diverse.

As curators of the healthcare hashtag project, we at Symplur have been especially impressed with the rapid rise in the number of patient initiated discussions taking place.  These discussions include people sharing insights into their diseases, treatments, recovery and more.  However, in an effort to provide the most useful tool that we can to help facilitate these discussions, we’ve always turned to the healthcare community itself to provide us with feedback, suggestions, and certainly … with hashtags.

With the above in mind, my colleague, Audun Utengen (@audvin) and I (@tmlfox) were thrilled and hono(u)red with the opportunity to not only participate in the #hcsmca tweetchat this past August 29, but were also given the privilege of submitting two questions for crowdsourcing with this remarkably active and insightful group.  Those questions were:

  1. From a patient perspective, are analytics like Top Influencers helpful as discovery tool in their disease experience?
  2. Do you see an ethical problem with companies mining people’s public tweets for analytical purposes? Where should the line be drawn?

What took place during the course of this segment of the chat was fantastic!  So much so that Audun and I wanted to follow up with our thanks … and with a bit of a recap of what we see as the “take-aways” were from the discussion.

Analysis of all mentions during August 29 #hcmsca chat

Centrality analysis of all mentions during August 29 #hcmsca chat

Do “Top Influencer Lists” matter to patients on Twitter?

It became clear at the onset that we had a very helpful mix of participants.  Patients, patient advocates, providers, and even parties who are personally familiar with and passionate about analytics.

On the issue of how helpful analytics such as “Top Influencers” are, there were some decidedly mixed reviews.  Some shared their personal experience using that feature of the project as a means to find people who had information or connections, and to not feel alone.

Others were rightfully quick to ask how “Influencers” are determined, pointing out that the “loudest” (most frequent tweets) do not imply reliable resources or influence.  And while we pointed out that we use the number of “@ mentions” and not number of “tweets” using a hashtag to measure influence, it was still pointed out that a single metric is not a good measure of true influence.

Our take-away on this part of the discussion is that we are in the early stages of identifying those individuals who we can recommend be followed based on an algorithm.  And we certainly have more work to do in this area.  However, we also attempted to very clearly point out that the healthcare hashtag project doesn’t attempt to “rank” people in the same manner as other sites, like Klout.  Rather, our attempt is to help new users to explore and to discover both people and conversations.

In the end, I think Annette McKinnon (@anetto) hit the nail on the head during the tail end of this segment of the chat when she said, “I think it is more introductory. You soon decide who you will keep watching and listening to”.

What are the ethical issues of data mining healthcare tweets?

Concerning the ethics of data mining people’s public tweets from these healthcare conversations, there was no shortage of opinion there either.

First off, there was considerable discussion about the “open discussions” on Twitter vs. the more “private discussions” in some forums.  And while it was pointed out that many forums are not private and only required signing up if one chose to contribute, others felt that Twitter was the most open platform on which to join in on these discussions.  Furthermore, we were surprised to learn from Aurelia Cotta (@AureliaCotta) that, “… there already are more pts on twitter, most I know never use hashtags and some misspell their disease to keep private/anon”.

While some expressed concern over “who” was data mining the healthcare tweets, there seemed to be a little less concern over “what” they were going to do with the information.  Some chimed in suggesting that Twitter itself allows one to keep their tweets private.  And if they don’t wish for their tweets to be public information, then they should reset their chosen privacy option.  I myself suggested that, “…Perhaps *not* analyzing these open discussions is in its own way ‘unethical’?”  After all, if done ethically, the richness of the health related data available has enormous potential to benefit those in need if this data gets into the right hands.

Our short term take-away is that this is the topic which probably has further reaching consequences and concerns as time goes on and more and more people share information on these open social media platforms.  There seems to be a desire for “reasonable” privacy even when conversations are completely public.  We believe that efforts by data miners to adhere to this desire is not only an ethical issue, but it’s an issue that if not addressed with care and genuine concern, could result in these conversations rushing to those closed, private portions of the web where there will be little or no discovery available to the masses.

A sincere “Thank You” to the #hcsmca community

I’d again like to say that both Audun and I are very grateful for the amazing opportunity that we had to pose these questions to #hscmca.  And we’re even more grateful for the wealth of candid insight that was so openly shared with us.

When it comes to the healthcare hashtag project, we at Symplur work hard at providing a meaningful and useful experience for both the novice and veteran user of healthcare social media.  But it’s like we’ve said from the beginning, it’s a “social project”.  One that, at its core, relies on interested individuals to make it a success for the healthcare community as a whole.

7 Comments leave one →
  1. September 7, 2012 1:59 pm

    This is a nice summary of the chat, which I participated in.

    I am curious about the image used in the post, “Centrality analysis of all mentions during August 29 #hcmsca chat”. Can you explain what is meant by centrality analysis in this context and what the overall message or pieces of information this image is attempting to convey? Thanks.

    Laura

    Like

  2. September 8, 2012 8:52 am

    Good question Laura. I’m going to bring Tom in to the conversation to explain the graphic.
    Colleen

    Like

  3. September 9, 2012 11:41 am

    Hi Laura,
    We strive very much to find ways to make it simpler to read the vast data in our database. As many other have, we find that visualizing data is the most efficient way. Tables and tables of data can be quite taxing.

    What you see is a visual representation of all the tweets from the #hcsmca chat on the August 29th. However, only tweets where a person was referenced (mention) is actually shown.
    – A mention created a link between these nodes. A thinker link indicates a stronger relationship which in this case means a richer conversation with multiple mentions.
    – The size of a node indicates Betweenness Centrality which indicates the importance of a node within the conversation. It answers the question of how central a person is in this network. It can be found to be more meaningful than just a simple mention ranking.
    – The color coding of the nodes is identifying communities within the network (Louvain method). It’s an algorithm visualizing sub-groups or clusters, and in this case I think it did a really good job. Although I’m not familiar with all the people that participated in the chat, but it seems to me that the algorithm did make some distinction between newcomers to the chat (symplur et al in green) and the regulars. However, that is not what it is really trying to say, but rather who formed a discussion group within the tweetchat itself.

    Hope this helps! :)

    Audun

    Like

  4. September 10, 2012 6:56 pm

    Hi Audun,

    Thanks for clarifying the content of the image.

    What I find of interest about your post is what you chose to discuss in your summary, in particular the tweets by Annette McKinnon and Aurelia Cotta, which you found valuable enough to quote directly. How would we find these using the centrality analysis as depicted in the image? How would we know to consider following Annette or Aurelia on Twitter if you had not shared the content of their tweets?

    If we look closely at the image you included we can find Annette and Aurelia’s Twitter handles in the visualization. I think this reveals a very important point about social media (and in particular Twitter) analytics. Yes, centrality analysis is interesting but it does not tell the whole story. Social network analysis (SNA) quantifies behaviour but do not qualify it. It is the content of the tweet that includes that experiential or anecdotal tidbit from a patient or a suggestion from a health professional that draws us to use social media such as Twitter. Using SNA to describe the relationships does not necessarily help those who are trying to use Twitter to find relevant, credible content. According to the centrality analysis image based on this hcsmca chat this information was most likely to come from Colleen Young. But what if the one piece of information that could help someone treat a medical condition comes from a singular or series of tweets shared by someone who uses Twitter infrequently?

    I do not think “Top Influencers” is a helpful analytic, at least in the way I think you have intended its use – a means to find credible information. Further, your statement “Our take-away on this part of the discussion is that we are in the early stages of identifying those individuals who we can recommend be followed based on an algorithm. And we certainly have more work to do in this area.” contradicts your own summary IMHO. Why would you recommend following individuals? Should we be following the content (or the community) instead? Isn’t the purpose to find relevant and trustworthy information regardless of its source? We know that communities of practice are highly regarded in terms of their capacity to support mutual engagement, information sharing and the co-creation of new knowledge.

    In order to further address these issues I suggest the following questions to the health care Twitter community: How do you locate, discern and implement credible healthcare information for personal (or professional) use?

    I also suggest these challenges to the social media analytics community: How do we measure the effectiveness of communities that are sharing health information? How do we measure the quality, relevance and applicability of the content of tweets? How do we measure whether knowledge is being translated and implemented by using Twitter?

    I have also used SNA to explore online Twitter-based community chats, in particular the #hcsmca community. You may want to read my two blog posts on these efforts:

    Network analysis of the #hcsmca Twitter community: lurking as a form of legitimate peripheral participation? (http://ogrdy.ca/tVq0Ad )

    Content analysis of #hcsmca tweets: the importance of context in social media analytics (http://ogrdy.ca/uYkPqQ)

    I hope you find my comments to be of value.

    Laura

    Like

Trackbacks

  1. Crowdsourcing healthcare social media analytics with #hcsmca ... | Healthcare Social Media News | Scoop.it
  2. Crowdsourcing healthcare social media analytics with #hcsmca ... | Health and Social Media | Scoop.it
  3. Crowdsourcing healthcare social media analytics with #hcsmca ... | #web2salute | Scoop.it

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: