Health Knowledge Sharing Network on Twitter

The article “Twitter hashtags for health: applying network and content analyses to understand the health knowledge sharing in a Twitter-based community of practice” investigates the health knowledge sharing by the community of practice (CoP) on social media platform- Twitter. The authors Xu et al (2014) utilize network and content analyses to examine the health-related conversations on Twitter. The researchers used centrality measures such as in-degree centrality, out-degree centrality and density measure K- core to analyze the Twitter data.

The article states that social media platforms like Twitter establish a community of practice (CoP) which enables peer collaboration and relationship development on the basis of shared identities, concerns, and interests. The authors argue that health communication via hashtags supports the basis of a virtual community of practice as it is driven by the shared interests and practice of knowledge sharing. In this study, they pose five separate research questions as listed below.

RQ1: What are the salient themes in health-related conversations via Twitter hashtags?

RQ2: Who are the central participants in the Twitter-based CoPs?

RQ3: What healthcare roles are central in the Twitter-based CoPs?

RQ4: Is the CoPs centralized and dominated by a few participants?

RQ5: What are the characteristics of the interactions between different healthcare roles?

Since RQ2, RQ3, and RQ4 are based on network analysis, I will focus on them in this blog.

Data collection:

The researchers used a python script to mine tweets associated with 14 specific hashtags that cover health-related conversation. The hashtags include #PTSD, #OCD, #BreastCancer, #HealthTalk, #LCSM, #LCAM2013, #BCSM, #Alzheimer’s, #BOTOX, #Addictionchat, #BrainTumorThursday, #TreatDiariesChat, #RDchat, and #HeartHealth. All the hashtags are associated with different medical conditions ranging from mental health conditions such as OCD and PTSD to heart, lung, and brain problems. The python script ran daily from October 25 to December 26 of 2013 and the raw dataset included 125,907 unique tweets. After downloading the tweets, researchers also documented the Twitter user profiles, the size of their followers, and their profile descriptions. To organize the collected data, they coded the information in different categories as listed below.

They also categorized the participants in according to their roles in healthcare conversations. The categories included Healthcare providers, Advocacy, Engaged Consumers, Average consumers, Media, Government, Non-health related organizations, and Unknown profiles.


To analyze the node’s position within network, centrality measure of in-degree, out-degree, and betweenness were used. Additionally, K-core was used to identify a cluster of densely connected participants. To address RQ4, they used the centralization and core/peripheral modeling to reveal network structure. The result suggests that the most central participant by in-degree is @EverydayHealth, a healthcare organization that specializes in delivering medical advice. Other top participants include practitioners, healthcare organizations, advocates and emphatic individuals. The results also suggest that same type of participants are also central by out-degree and betweenness centrality measures. A cluster of densely connected participants identified using K-core. The cluster includes five practitioners, four empathetic individuals, one researcher and one healthcare activism or advocacy organization.

A core/periphery network is determined as it means the presence of a densely connected core or a loosely connected periphery. The core/peripheral structure was determined by fitting the entire network to the core/peripheral model in UCINET. The final fitness of the conversation network to core/peripheral model is .262, which is below the .5 threshold. Overall, the conversation networks are decentralized, and the entire network exhibits no core/periphery structure (Xu et al, 2014).


Xu, W. W., Chiu, I., Chen, Y., & Mukherjee, T. (2014). Twitter hashtags for health: applying network and content analyses to understand the health knowledge sharing in a Twitter-based community of practice. Quality & Quantity, 49(4), 1361-1380. doi:10.1007/s11135-014-0051-6

One thought on “Health Knowledge Sharing Network on Twitter

  1. What is interesting about this article is that it is sophisticated methodologically but I don’t see any discussion of the implications of their findings. So what? Why should we care that @everyday health was highest? What are the implications of these findings? This is where digital sociology comes in! We can help computational social scientists figure out what their findings mean. We can work with FB engineers to help them understand what they are finding as well as the implications for how they are collecting and deploying data. Why would we want to know what type of organization is at the center of distributing health information on Twitter? What if the central organization was anti-science? Or religiously based? Or insurance company based? These are digital sociology questions!

Leave a Reply

Your email address will not be published. Required fields are marked *