Ever been in a national conference, heard some really important information, and wondered why the rest of the world wasn’t hearing the same thing?

Well, you’re not the only one.

Recently, researchers from East Carolina University, University of North Carolina Chapel Hill, and Carnegie Mellon assessed the use of Twitter by the American Society of Nephrology to inform and educate the public about kidney disease at its annual conference referred to as “Kidney Week.”

Their research article provides some useful information to medical conference organizers worldwide. So, tweet away!

Quick Background

The American Society of Nephrology began focusing on tweets to the public because it recognized that there was a tremendous lack of understanding about, and recognition of, kidney disease – despite the numerous people suffering from kidney problems in the US. The organization’s former president pointed out that public awareness and education are key challenges for the group in the coming years.

Other organizations face similar challenges as the public continues to receive an onslaught of information online, increasingly with mobile devices, that is not evidence-based. A key question for this and other groups is what is an effective health information communication tool – an old question for public health. Social media was viewed as one potentially effective tool. The researchers sought to determine its effectiveness.

Tweeting Hypothesis

In order to determine whether tweets were informing the public about kidney disease, the researchers hypothesized that content, citation, and sentiment analyses of tweets generated from Kidney Week 2011 would help them identify educational tweets being disseminated to the public.

The researchers perceived an ideal tweet as one which had informational content, internal citations, and positive sentiment score, i.e. people liked the information.

Sentiment analyses refer to techniques for assessing statements (usually online statements) of numerous individuals in order to determine how they feel about a particular subject.  For example, a business could use sentiment analysis to determine what people think about its new product.  Doctors could use it to determine what patients think about their services.  It’s sort of an overt way to spy on people and see what they think about something.

Methods for Assessing Tweets

Researchers analyzed tweets produced during Kidney Week 2011 from November 8th – 13th.  They focused on public tweets that referenced Kidney Week 2011. The timeline was available worldwide to anyone searching using the “hashtag#kidneywk11 using any Twitter client or the Twitter website, regardless of ASN membership status.

The ASN promoted it as a place to collect tweets about the meeting. The researchers de-identified all tweets prior to analysis and they made the assumption that tweeters using  #kidneywk11 participated in the conference.

Researchers analyzed the content of all tweets in the public timeline, labeling them as informative or uninformative based on an industry standard. They were further defined into more detailed categories regarding their information content. Tweets were also classified as having an internal, external, or no citation.

A “retweet” was referred to as an internal citation. They occur when a person resends a tweet verbatim to someone else – essentially forwarding a message. Sentiment scores for each tweet were calculating using a modified Affective Norms for English Words sentiment lexicon (AFINN)

Key Tweet Results

The researchers discovered that informational content was found in 29% of the Tweets sent out during Kidney Week. They noted that this was more than a similar conference in 2011, American Diabetes Association conference, which had 16%. There were significantly more internal cites than external (38% versus 22%, p<0.0001) of these tweets. This makes sense given that retweet takes less time than writing a new tweet.

The sentiment score for the informational tweets were significantly more negative than the scores of tweets without informational content (means −0.162 versus 0.199 respectively, p<0.0001). This indicates that people perceived more of these tweets in a negative way.

Given the complexity of analyzing the language of tweets, these negative scores could be anti-factual since they  may have been associated with words such as disparities, problems, or death and other terms that may be common informational content relevant to disease. A decrease in deaths, though, might be perceived as negative whereas it’s actually positive.

Future Tweeting Research?

The researchers note that their findings can inform future nephrology-focused conferences. Given that numerous medical specialties share their concerns of trying to keep the public well informed with reputable facts, their results should also inform the efforts of conferences focused on other diseases – especially those that have historically had strong patient communities-like breast cancer.

Some future questions for the researchers might be who actually has access to Twitter. Is it limited by age or income level? Also, are the tweets communicating information in a language that is user friendly to the public including translating some tweets into Spanish or any other languages that are relevant to the population of people with kidney disease in the US?