A neighborhood analysis of Twitter communication in emergency medicine

To the editor:

Over the last decade, social media and free open access medical education (FOAM) has emerged as popular forms of education and information dissemination within Emergency Medicine. Many clinicians, from trainees to board-certified physicians, increasingly interact and share information on social media platforms such as Twitter [1]. Twitter has been popular among medical educators as a platform for disseminating information [2]. Within Twitter, it is common practice for users to “tweet at” or mention users when posting, in order to share information. Typically, this information contains brief summaries of recent publications, interesting cases, and medical opinions. With Twitter, users attract followers that interact with other users. These relationships create “neighborhoods” of communication and interaction. Despite the growing role of social media in medical education, less is known regarding the extent and type of communication. The objective of this study was to analyze existing Twitter interactions and relationships among influential EM Twitter users. Such information would provide a broad overview of the connections and relationships within these new educational media.

We conducted a planned social network analysis of 71 of the most influential EM Twitter users, as determined by a series of variables including the number of followers, frequency of tweets, national recognition, and previous research in this area. The list was generated from a previous literature identifying the most influential users and updated by this paper’s authors, all emergency medicine physicians with Twitter presence [3]. None of the authors appeared on this list. Data was collected from a publicly-available database and received IRB exemption. The most recent 200 tweets of each individual user published online, as of August 2019, were collected from these 71 users. The data was then queried via a Standard Twitter® Application Programming Interface (API). All statistical analyses and model constructions were performed using a high-level, general-purpose programming language, Python, and its associated open statistical packages [4], while all tweets were analyzed with Natural Language Processing [5].

After parsing the text of the 71 EM users, 1191 unique Twitter users were mentioned. From this subset, private Twitter accounts were removed from analysis. A modularity score was calculated and each user was classified into an individual neighborhood. Table 1 identifies the top 30 out of a total 1191 EM Twitter users with the highest centrality degree scores. Fig. 1 demonstrates the general overview of the several regional neighborhoods of Twitter users. These neighborhoods, demarcated by different colors, suggest the relationships between its users are close. Fig. 1 again demonstrates 16 different “neighborhoods” as defined by the modularity categories within the EM community, which is displayed in the table. The majority of the users identified with a single neighborhood. However, several Twitter users appear to interact with multiple neighborhoods.

Table 1.

Top 30 Twitter users arranged by highest degree centrality scores.

Twitter user	Modularity class	In-degree	Out-degree	Total degree
@EMHighAK	8	815	61	876
@mdaware	5	37	605	642
@painfreeed	10	363	168	531
@tchanmd	7	23	503	526
@EMNerd_	4	398	100	498
@pemedpodcast	15	304	158	462
@rogerrdharris	0	275	184	459
@TheSGEM	6	373	86	459
@emcrit	4	44	374	418
@srrezaie	10	38	371	409
@ketaminh	13	68	291	359
@lwestafer	5	178	158	336
@broomedocs	15	77	244	321
@brent_thoma	2	42	278	320
@smithecgblog	15	99	205	304
@EMO_Daddy	2	116	185	301
@emswami	10	82	215	297
@meganranney	6	33	258	291
@emergidoc	9	144	146	290
@m_lin	7	87	200	287
@rcempresident	0	59	220	279
@richardbody	0	92	182	274
@emlitofnote	5	114	159	273
@em_educator	7	46	226	272
@_nmay	0	24	244	268
@socraticem	3	106	162	268
@drjessepines	14	83	185	268
@pedemmorsels	15	87	174	261
@cabreraerdr	10	42	218	260

Fig. 1.

The overview of the different “neighborhoods”.

Our analysis identified a vast social network connected to over 1000 other Twitter users. As online platforms have risen, many emergency physicians have turned to social media to social media to share information in a digestible and rapidly transmitted format. Our findings demonstrate a means whereby one user can reach a large, diverse audience both directly and indirectly connected. EM encompasses a large array of topics ranging from trauma to psychiatric emergencies [6,7], and it is not surprising that different niches have developed within the EM community. Numerous areas of future study within education research and learning methodology arise from this network analysis. For example, one may evaluate whether large social media influence translates into actual changes in practice or clinician behavior. Additionally, a broader question of the assessment of brief social media tweets with regards to scientific accuracy and rigor remains an area ripe for investigation. For the most part, unlike more traditional peer-reviewed academic publications, social media messaging is largely unverified, with limited space to delve deeply into methodological strengths and weaknesses of a particular study. Examination of not just the reach and influence of social media users, but also the quality of evidence disseminated may provide novel insight into how such innovative methods may complement existing means of scientific inquiry.

Our study had several limitations. This study focused exclusively on Twitter. Many EM Twitter users use other platforms such as Facebook, Research Gate, and podcasts, which were not accounted in our model. Furthermore, the data examined English-speaking Twitter users and may have not included influential non-English speaking users.

Medical education using social media has seen an explosion of interest with regard to information dissemination with platforms, such as Twitter. Our results provide information regarding social media interactions within the EM community. The use of such graphical and numerical representation provides a novel means of analyzing social media use and its impact throughout the EM social media community. Our findings could inform future studies examining the use and functionality of such platforms in medical education.