Comparison of Intercom and Megaphone Hashtags Using Four Years of Tweets From the Top 44 Schools of Nursing: Thematic Analysis

Background When this study began in 2018, I sought to determine the extent to which the top 50 schools of nursing were using hashtags that could attract attention from journalists on Twitter. In December 2020, the timeframe was expanded to encompass 2 more years of data, and an analysis was conducted of the types of hashtags used. Objective The study attempted to answer the following question: to what extent are top-ranked schools of nursing using hashtags that could attract attention from journalists, policy makers, and the public on Twitter? Methods In February 2018, 47 of the top 50 schools of nursing had public Twitter accounts. The most recent 3200 tweets were extracted from each account and analyzed. There were 31,762 tweets in the time period covered (September 29, 2016, through February 22, 2018). After 13,429 retweets were excluded, 18,333 tweets remained. In December 2020, 44 of the original 47 schools of nursing still had public Twitter accounts under the same name used in the first phase of the study. Three accounts that were no longer active were removed from the 2016-2018 data set, resulting in 16,939 tweets from 44 schools of nursing. The Twitter data for the 44 schools of nursing were obtained for the time period covered in the second phase of the study (February 23, 2018, through December 13, 2020), and the most recent 3200 tweets were extracted from each of the accounts. On excluding retweets, there were 40,368 tweets in the 2018-2020 data set. The 2016-2018 data set containing 16,939 tweets was merged with the 2018-2020 data set containing 40,368 tweets, resulting in 57,307 tweets in the 2016-2020 data set. Results Each hashtag used 100 times or more in the 2016-2020 data set was categorized as one of the following seven types: nursing, school, conference or tweet chat, health, illness/disease/condition, population, and something else. These types were then broken down into the following two categories: intercom hashtags and megaphone hashtags. Approximately 83% of the time, schools of nursing used intercom hashtags (inward-facing hashtags focused on in-group discussion within and about the profession). Schools of nursing rarely used outward-facing megaphone hashtags. There was no discernible shift in the way that schools of nursing used hashtags after the publication of The Woodhull Study Revisited. Conclusions Top schools of nursing use hashtags more like intercoms to communicate with other nurses rather than megaphones to invite attention from journalists, policy makers, and the public. If schools of nursing want the media to showcase their faculty members as experts, they need to increase their use of megaphone hashtags to connect the work of their faculty with topics of interest to the public.


Introduction
Twitter is a microblogging website where users can post "tweets" (brief messages, images, and videos) to share with "followers" (people who have chosen to follow their Twitter account). Hashtags are words or phrases (without spaces) that are preceded by a pound sign (#) [1]. Hashtags first came into use on Twitter in 2007 when a user named Chris Messina put forward a proposal for "…improving contextualization, content filtering, and exploratory serendipity within Twitter" [2]. In his proposal, Messina wrote that his primary interest was "simply having a better eavesdropping experience on Twitter" [2]. In 2018, hashtags were widely used on Twitter to make tweets easy to find for other Twitter users interested in a given topic.
When the landmark Woodhull Study on Nursing and the Media was published in 1998, the voices and faces of nurses were found to be largely absent from news stories [3]. Mary Chaffee wrote that "[t]his lack of visibility limits nursing's ability to communicate important health information, impedes nursing's ability to define its role and contributions in the health care delivery system, and restricts nursing's ability to advocate for health policy" [4]. Because Twitter was not launched until 8 years after the Woodhull Study was conducted, the researchers obviously could not look at Twitter data in their analysis. Shattell and Darmoc argue that nurses should consider using Twitter to make their "practical, real-life knowledge or…research findings or insights on current issues… available for the public" and to "harness attention from some more traditional media sources" [5]. While there is an abundance of research regarding the use of hashtags by health care professionals on Twitter [6][7][8][9][10], little is known about the ways in which schools of nursing used Twitter to invite attention from and engagement with journalists, policy makers, and the general public in the 2 years before The Woodhull Study Revisited was published in September 2018 and the 2 years after its publication. This study seeks to fill this gap.
When this study began in 2018 as a last-minute addition to The Woodhull Study Revisited, I sought to determine the extent to which the top 50 schools of nursing were using hashtags that could attract/invite attention from journalists on Twitter [11]. Preliminary findings using 2016-2018 data were intriguing but were not published with the rest of the results of The Woodhull Study Revisited [12]. In December 2020, the timeframe was expanded to encompass 2 more years of data so that before and after Woodhull Study Revisited analyses could be conducted. In addition, the scope was expanded to include an in-depth analysis of the types of hashtags used by schools of nursing. The resulting study is a comprehensive analysis of 4 years of tweets from the top 44 schools of nursing in the United States.
Methods have been described in detail using plain language so that researchers can easily replicate the study without needing specialized knowledge in natural language processing or machine learning. Democratizing Twitter analysis requires greater transparency regarding the methods used. As such, each table in this manuscript illustrates a step in the data analysis process that would otherwise be opaque to readers if the step was simply described in the narrative.

Research Question
The study sought to answer the following question: to what extent are top-ranked schools of nursing using hashtags that could attract/invite attention from journalists, policy makers, and the general public on Twitter? Below is a detailed description of the methods used for sampling, data collection, and data analysis.

Sampling
When this study began in February 2018, the sample of nursing schools was drawn from US News and World Report's 2017 list of the top nursing schools with master's degree programs. Fifty of the highest-ranked schools were selected from this list, with numerical rankings ranging from 1st to 48th (with several ties). The US News and World Report rankings were used as a mechanism for identifying the schools of nursing to include in this study with the knowledge that the rankings do not necessarily mean that the schools included at the top of the list are inherently "better" than the schools ranked lower. The decision to include the 50 highest-ranked schools of nursing in the sample was based on the fact that the US News and World Report rankings are the primary way that members of the media can quickly identify top schools of nursing nationally. The US News and World Report gets 7 million unique visitors to the education rankings and information webpages each month (US News and World Report, 2018).
In February 2018, of US News and World Report's 50 top schools of nursing, two schools did not have a Twitter account and one school had a locked private Twitter account that was inaccessible to anyone other than those who were given permission by the school to follow the account. Thus, the school of nursing with the locked Twitter account and the two schools without a Twitter account were excluded from the 2016-2018 data set. The three schools omitted from the 2016-2018 data set are indicated in Table 1. In December 2020, when the second phase of this study was conducted, 44 of the original 47 schools of nursing still had public Twitter accounts under the same name used in the 2016-2018 data set. The three schools that no longer had a public Twitter account under the same name in 2020 are indicated in Table 1 and were omitted from both the 2016-2018 and 2018-2020 data sets for the sake of consistency.

Data Collection
Data collection was conducted twice during this study. In February 2018, a list of the top 50 schools of nursing was matched with publicly accessible Twitter accounts and then a data request was submitted to Export Tweet for the most recent 3200 tweets from each of the top-ranked schools of nursing. Because schools of nursing tweet with varying frequency, the past 3200 tweets for any given school of nursing covered a wide array of time frames. At one end of the spectrum, there were five schools of nursing, including Vanderbilt University, Johns Hopkins University, University of Michigan-Ann Arbor, Boston College, and University of Pennsylvania, for whom the oldest tweet in the data set was from 2016. At the other end of the spectrum, there were five schools of nursing, including University of Virginia, Yale University, Case Western Reserve University, University of Utah, and University of North Carolina-Chapel Hill, for whom the oldest tweet was from early 2009. Table 2 lists the oldest tweet in the data set from each school, with schools of nursing listed in order of their oldest tweet in the data set.  Table 2 was used to determine the most recent "oldest tweet" date in the 2016-2018 data set. The @VanderbiltNurse Twitter account had the most recent "oldest tweet" (September 29, 2016), so September 29, 2016, was selected as the start date for the analysis. This meant that the time period to be covered in the 2016-2018 data set would be September 29, 2016, through February 22, 2018. Tweets with dates older than September 29, 2016, were filtered out from the data set, resulting in 16,939 tweets for the 2016-2018 data set. Table 3 describes the composition of the final 2016-2018 data set, with schools listed in alphabetical order by Twitter account name. During phase two of the study, a data request was submitted to Vicinitas for all tweets from February 23, 2018, through December 13, 2020, from the 44 still-active Twitter accounts.
Tweets prior to February 23, 2018, were deleted from the data set. Table 4 lists the oldest tweet in the 2018-2020 data set from each school, along with the number of tweets per school. After cleaning the data, the 2016-2018 and 2018-2020 data sets were merged into a single data set containing 57,307 tweets. Table 5 describes the composition of the new 2016-2020 data set, with schools listed in alphabetical order by Twitter account name.
In December 2020, the original list of 47 schools of nursing was matched with publicly accessible Twitter accounts. Of the original 47 schools of nursing, 44 still had public Twitter accounts under the same name used in the first part of the study.
The three Twitter accounts that were no longer active (@UICollegeofNurs, @UACON, and @CU_Nursing) were removed from the original data set, resulting in a data set containing 16,939 tweets from 44 top-ranked schools of nursing. The most recent 3200 tweets from each of the Twitter accounts were extracted and analyzed. Excluding retweets, there were 40,368 tweets for the time period covered (February 23, 2018, through December 13, 2020). These 40,368 tweets were added to the data set, resulting in a data set containing 57,307 tweets from September 29, 2016, through December 13, 2020.

Data Analysis
The analyses in this study were conducted using R version 4.0.3 (Bunny-Wunnies Freak Out), R Studio Version 1.3.1093, and Microsoft Excel for Mac Version 16.43. The following are the steps taken to generate a list of the most frequently used hashtags in the 2016-2020 data set, along with the number of times each hashtag appeared. Initially, the Excel file was uploaded to R software. The R Markdown package was installed, and the elements of Van Horn and Beveridge coding were used [13]. The text strings in the data set were cleaned. The character encoding in tweets was homogenized to remove the strings of nonsense characters indicating the presence of emojis in the source tweets. This converted character encoding to Unicode UTF-8. Thereafter, capitalization in tweets was removed by turning everything into lowercase. Subsequently, extra whitespace and URLs were removed from the tweets. Once the text strings were cleaned, the hashtags present in the data set were identified and a list of the hashtags from most to least frequently used was generated. The data frame generated in R was exported to Excel, with hashtags listed in one column and their frequency in another. The corresponding script in R has been provided in Multimedia Appendix 1 so that readers can replicate the analysis.
Because there was interest in detecting changes in the use of hashtags by schools of nursing after the results of The Woodhull Study Revisited were published in Fall 2018, the steps described above were repeated to split the 2016-2018 data set into two parts. The first covered September 29, 2016, through September 27, 2018 (the day that The Woodhull Study Revisited was published in the Journal of Nursing Scholarship), and the second covered September 28, 2018, through December 13, 2020. The same process outlined previously was used to analyze the data and generate frequency tables for the hashtags used during each time period of interest.

Results
There were 6866 different hashtags used in the 2016-2020 data set. All hashtags that had been used 100 times or more across the entire corpus of tweets in the data set were identified, and these 71 hashtags were characterized as being those with the highest frequency of use by the schools of nursing in the study. These 71 hashtags were used a total of 26,243 times in the 2016-2020 data set, as detailed in Table 6. Among the 6866 different hashtags appearing in the 2016-2020 data set, 3774 were used only once and 6178 were used 10 or fewer times. #veterans 101 #uofunursing

Typology of Frequently Used Hashtags
Using Excel, a thematic analysis was conducted of the hashtags that were used 100 times or more in the 2016-2020 data set. Collectively, the 71 hashtags were used a total of 26,243 times. To conduct the thematic analysis, the list of 71 frequently used hashtags was considered and similarities were assessed. As similarities were identified, the hashtags were grouped into categories, and this process of coding (and recoding) hashtags was continued until there were six categories that explained the vast majority of the hashtags. A seventh category was added to capture the assortment of hashtags that did not lend themselves to categorization. The following seven types of hashtags emerged during the process of thematic analysis: (1) Nursing, hashtags about nurses, nursing, nursing degrees, nursing licenses, etc; (2) Schools, hashtags about universities, schools, colleges, mascots, or locations; (3) Illness/disease/condition, hashtags about illnesses, diseases, conditions, or awareness day/month; (4) Population, hashtags about populations that nurses serve; (5) Health, hashtags about health care, health, global health, etc; (6) Conference or tweet chat, hashtags about conferences or specific Twitter chats for health care professionals; (7) Something else, hashtags that did not fit into one of the other six categories. Table 9 lists the hashtags contained in each of the seven categories. For the purposes of this study, the seven types of hashtags were considered to be either inward facing ("intercom hashtags") or outward facing ("megaphone hashtags"). Intercom hashtags were those intended to invite attention from/interaction with nurses, members of the university/school community, or attendees at a nursing conference or Twitter chat. Megaphone hashtags were those intended to invite attention from/interaction with people such as journalists, policymakers, and the general public.
The intercom hashtag types were as follows: nursing (hashtags about nurses, nursing, nursing degrees, nursing licenses, etc); schools (hashtags about universities, schools, colleges, mascots, or locations); and conference or tweet chat (hashtags about conferences or specific Twitter chats for health care professionals). The megaphone hashtag types were as follows: illness/disease/condition (hashtags about illnesses, diseases, conditions, or awareness day/month); population (hashtags about populations that nurses serve); health (hashtags about health care, health, global health, etc); and something else (hashtags that did not fit into one of the other six categories).
The vast majority of the 71 hashtags that were used 100 times or more in the 2016-2020 data set can be categorized as intercom hashtags (inward-facing hashtags focused on in-group discussion within and about the profession). Collectively, nursing hashtags (n=9810, 37.4%), school hashtags (n=10,974, 41.8%), and conference or tweet chat hashtags (n=1003, 3.8%) comprised 83.0% (n=21,787) of the 26,243 times that the 71 frequently used hashtags occurred in the data set.
In contrast, few of the 71 hashtags that were used 100 times or more in the 2016-2020 data set can be categorized as megaphone hashtags. Collectively, health hashtags (n=1801, 6.9%), illness/disease/condition hashtags (n=1211, 4.6%), and population hashtags (n=170, 0.7%) comprised 12.1% (n=3182) of the 26,243 times that the 71 frequently used hashtags occurred in the data set. When the "something else" hashtags (5%) were added, the total of megaphone hashtags was approximately 18% of the 26,243 times that the 71 frequently used hashtags occurred in the data set.
When the data set was divided into two parts to detect changes in the use of hashtags by schools of nursing after the results of The Woodhull Study Revisited were published, the findings were similar to those of the analysis of the data set as a whole, with one notable exception. Prior to the publication of The Woodhull Study Revisited on September 27, 2018, none of the hashtags that were used 100 times or more pertained to an illness, disease, or condition. In the 2 years after the publication of The Woodhull Study Revisited, 7% of the frequently used hashtags pertained to an illness, disease, or condition. Further analysis revealed that this shift was attributable to the use of the following two hashtags: #covid19 (n=895) and #hiv (n=115).

Missed Opportunities for Tweeting About Trending Topics
Of the 6866 different hashtags appearing in the 2016-2020 data set, 6178 were used 10 times or less. These seldom-used hashtags included a number of hashtags that were widely used on Twitter during the time period covered by this study. Table  10

Discussion
Although the top 44 schools of nursing have an active social media presence on Twitter, collectively, their use of hashtags functions more like an intercom to communicate with other nurses rather than a megaphone to invite attention from and dialogue with journalists, policy makers, and the general public. Because intercom hashtags are both inward facing and overused, they are of minimal use when it comes to drawing attention from and interacting with people outside of nursing. If schools of nursing want the media to showcase the voices of their faculty members as experts, schools of nursing need to be more strategic in their use of hashtags on Twitter. In order to accomplish this, schools of nursing need to increase their use of megaphone hashtags to connect the work of their faculty and students with topics and events of interest to the general public. For example, when topics like #guncontrol are trending, schools of nursing could tweet about the work their faculty members are doing in violence prevention.
On Twitter, schools of nursing have a unique opportunity to amplify the voices of their faculty members on health-related topics of widespread public interest like the impact of systemic racism on health, gun violence, and access to care, among others. If schools of nursing continue to use mostly intercom hashtags on Twitter, they will have squandered a powerful opportunity to share their expertise beyond the boundaries of the discipline.