#4645Boricuas: Twitter Reactions to the Estimates of Deaths by Hurricane María in Puerto Rico

Abstract

Aims:

This study explores the sentiment expressed by Twitter users after the Harvard University report was released, reporting a death estimate of 4,645 Puerto Ricans following Hurricane María.

Method:

Researchers utilized the NVivo addition NCapture to collect Twitter data including the hashtag #4645Boricuas. Thematic content analyses explored emergent themes within the hashtag. Geographic information systems (GIS) documented the location of Tweets and differences across geographical locations among Twitter users.

Results:

The themes that emerged in #4645Boricuas included: disaster phases, media, psychological processes, and politics. GIS documented that Tweets from Puerto Rico were more likely to discuss themes of psychological processes, politics, and resilience compared with disaster phases or media.

Conclusion:

The present data highlights how individuals engage with Twitter to cope following the trauma of natural disasters. In the aftermath of Hurricane María, individuals utilized Twitter to express their disillusionment with the government response to the disaster.

The United States (USA) has experienced several severe natural disasters in the past 15 years, including Hurricanes Katrina in 2005, Sandy in 2012, and Harvey and Irma in 2017. The impact of such natural disasters are measured in terms of both acute and immediate impacts, including death toll, the number of injuries sustained, destruction to property and infrastructure (Green, 1991), and incurred economic loss (Toya & Skidmore, 2007). The psychological consequences of natural disasters can also persist long-term; including impacts on life satisfaction, stress, health, physical symptoms, depression (Tremblay, Blanchard, Pelletier, & Vallerand, 2006), and posttraumatic stress disorder (PTSD; Spurrell & McFarlane, 1993).

There are numerous factors that influence how individuals respond to a natural disaster. Individuals who experience greater property loss during disasters tend to experience poorer health outcomes in the aftermath (Murphy, 1986). Marginalized populations are also at greater risk of experiencing poor outcomes following natural disasters (Davidson & McFarlane, 2006; Elliott & Pais, 2006). Additionally, the capacity of local governments to organize across institutional, human resource, policy, financial, technical, and leadership spheres may significantly mitigate the harmful outcomes of natural disasters and improve community success in post-disaster recovery (Kusumasari, Alam, & Siddiqui, 2010). Unfortunately, following a natural disaster, public desire for accountability can create focus on identifying individuals to hold accountable for the losses sustained by a community (i.e., “scapegoating”), an approach that overlooks inherent system weaknesses that contribute to poor responses to natural disasters, and fosters defensiveness among community members and leaders (Drabek, 1994), which can sabotage effective reconstruction efforts.

Emotional responses to a natural disaster also vary over time. Individuals coping with difficult emotions following a natural disaster may navigate multiple stages of grief (Kubler-Ross, 1973), including denial, anger, bargaining, depression and acceptance. According to DeWolfe’s (2000) phases of disaster model, individuals tend to experience emotional highs and lows across the phases of predisaster (i.e., warnings and threats), immediate impact, acts of heroism, a “honeymoon” period, a “disillusionment” period (especially salient around the anniversary of a disaster), and a “reconstruction” period, which includes working through grief, coming to terms with the loss, and experiences of setback. Greater coping skills may serve as protective factors against distress and worsened symptomatology following natural disasters (Benight & Harper, 2002). As described in the Ways of Coping Questionnaire (WCQ) (Folkman & Lazarus 1985, 1988; Folkman, Lazarus, Dunkel-Schetter, DeLongis, & Gruen, 1986), coping strategies that may assist in reducing distress after a traumatic event include confrontation, distancing, self-controlling responses, seeking social support, accepting responsibility, escape-avoidance, planful problem-solving, and positive reappraisal.

Social media can be a source for coping (e.g., blogs, social networks; Sutton, Palen, & Shklovski, 2008), and it has been described as a significant and valid form of communication for disasters (Palen, 2008). As people use social media to communicate in disasters, a framework of disaster social media use developed by Houston et al. (2015) consists of three disaster phases: pre-event, event, and post-event. Uses for the phases of event and post-event include: documentation, sharing and learning news, providing and receiving disaster response, creating awareness, providing and receiving support, expressing emotions and memorializing victims, telling and hearing stories, discussion of socio-political causes, and responsibility of events, connecting with community members (Houston et al., 2015). In addition to these uses, information and communication technology that includes web-based social networking (e.g., Twitter) helps people to find community during disasters (e.g., wildfires, hurricanes), including for people who have been displaced after disasters, and to create resources for communities (Shklovski, Palen, & Sutton, 2008).

Context of Hurricane María

In September of 2017, Hurricane María, a category four hurricane, hit the USA Commonwealth of Puerto Rico (P.R.). The hurricane damaged and destroyed roads and bridges, and interrupted electricity, water supply, and communication networks (Campbell, 2018). Additionally, the hurricane brought a high magnitude of disaster exposure to individuals in P.R. (Zorilla, 2017). A survey of over 96,000 youth found that 84% and 46% of school students saw houses damaged or had damage to their own home, respectively, 32% did not have enough food or water following the disaster, 30% believed their lives to be at risk, and 58% had a friend or family member leave P.R. (Orengo-Aguayo, Stewart, de Arellano, Suárez-Kindy, & Young, 2019). The research estimated that an additional 114,000–213,000 Puerto Ricans would leave the island each year following the disaster (Melendez & Hinojosa, 2017).

Whereas the official death toll estimated by the government of P.R. was 64, a later report by Kishore et al. (2018) estimated that 4,645 Puerto Ricans died as an aftermath of Hurricane María. A subsequent independent report commissioned by the Government of P.R. estimated a total post-hurricane mortality of 2,975 deaths (George Washington University, 2018), making Hurricane María one of the deadliest hurricanes in USA history. Importantly, the Jones-Shafroth Act of 1918 established P.R. as a Commonwealth of the USA. Puerto Ricans are American citizens, due federal aid and support following natural disasters. Whereas community-level organizing played a significant role in responding to the disaster, organizing efforts and political pressure from Puerto Ricans not living on the island also played a role in facilitating a response to the emergency and sustaining a longer-term public dialogue regarding the crisis (see Rodríguez-Díaz, 2018).

Following the release of the Kishore et al. (2018) report, individuals within P.R. and off the island took to Twitter to express their reactions to the death toll of Hurricane María via the hashtag #4645Boricuas. Hashtags are utilized to connect a Tweet to a larger public dialogue on the social media platform, and allow Twitter users to engage in community formation, as well as “hashtag activism” (Yang, 2016), defined as collective social media protest that involves raising public awareness and visibility around a public issue that might be otherwise neglected by traditional media outlets. #4645Boricuas was intended to draw attention to the 4,645 deaths of Boricuas, a demonym for the people of P.R. Extant literature has analyzed social media use in the aftermath of natural disasters, and has established the utility of examining public reactions to natural disasters via Twitter to assess community-level mental health, measure reactions to political and institutional figures, and clarify resource needs following natural disasters (Gruebner et al., 2017; Rhodes & Tran, 2012).

The Current Study

Recently, literature has highlighted the utility of social media to plan community responses to an impending natural disaster, provide crisis information in real-time to individuals impacted by a natural disaster, and improve stakeholder understanding of community experiences following a natural disaster (Houston et al., 2015; Velev & Zlateva, 2012). By promoting communication throughout various stages of a disaster, the use of social media may help mitigate negative outcomes for individuals and communities (Dufty, 2012; Gao, Barbier, & Goolsby, 2011; Wiederhold, 2013). Accordingly, the present study aims to explore how individuals in P.R. versus the mainland USA utilized the Twitter hashtag #4645Boricuas to respond to news regarding the Hurricane María death toll. Informed by the theory describing the phasic nature of community- and individual-level coping with disasters (DeWolfe, 2000) and theory describing the stages of grief (Kubler-Ross, 1973), thematic qualitative coding of Tweets with the hashtag #4645Boricuas sought to gain a deeper understanding of the psychological processes of individuals who participated in public dialogue regarding Hurricane María, and document public perception of the government response to the natural disaster. Researchers hypothesize that individuals tweeting from within P.R. and in the mainland USA would express similar sentiments regarding the impact of the disaster and the governmental response, given the large number of Puerto Rican living in the mainland USA.

Methods

On May 29, 2018, the Kishore et al. (2018) article estimating that 4,645 deaths had been either directly or indirectly caused by Hurricane María was published. This new death toll made international headlines and was immediately picked up by popular national news outlets, including the Washington Post and New York Times. Shortly thereafter, #4645Boricuas “trended” (i.e., was widely used) on the social media website Twitter. The research team utilized the NVivo addition NCapture to capture and download Tweets including the hashtag. Consistent with existing literature on Twitter analysis (Bogen, Millman, Huntington & Orchowski, 2018; Bogen, Bleiweiss, & Orchowski, 2019; Lachmar, Wittenborn, Bogen, & McCauley, 2017), Tweets were collected over 5 consecutive weekdays, from May 30th, 2018 to June 6th, 2018, to follow the course of online discourse during the peak of the #4645Boricuas “trend.”

To account for variability in online traffic and to ensure data was being collected from similar user samples, Tweets were collected at approximately 11 am each weekday. The original dataset included 33,549 Tweets. To ensure analytic focus on original user content, researchers removed all retweets (e.g., Tweets that may express identical sentiment between user) from the dataset. Research focused on original content is perceived as a more egalitarian qualitative research model, as including retweets within a larger dataset may disproportionately magnify the perspectives of celebrities and individuals with greater social media influence. A separate dataset of the five “top” Tweets (i.e., most frequently retweeted) were analyzed separately from the original content dataset to assess a smaller sample of high-impact Tweet content.

Tweets were excluded if they were not written in either English or Spanish, if their content was not associated with Hurricane María or the 4,645 deaths associated with the Hurricane, or if they simply included a website link or series of hashtags with no additional commentary. Although geo-localized Tweets constitute a small percentage of overall Tweets, this study was specifically designed to assess whether qualitative thematic differences arose across the mainland USA versus the island of P.R. As such, Tweets were removed from the dataset if they were not associated with specific latitude and longitude data. Twitter provides the following information on how latitude and longitude points are assigned to Tweets within their application programming interface: “If an exact location is provided (by the user), the coordinates object will provide a (longitude, latitude) array with the geographical coordinates, and a Twitter place that corresponds to that location will be assigned” (Twitter, 2019). To a location to be assigned, Twitter users must, therefore, assign a specific location, or have their autolocator enabled, when they post a Tweet.

After being reviewed for Tweet content relevance and geographic data, 630 Tweets were included in the final dataset. This final N is consistent with other qualitative thematic analyses of Twitter data (Bogen, Bleiwess, & Orchowski, 2019; Bogen, Millman, et al., 2018). In keeping with the best practices regarding Twitter data analysis and media user privacy protections, the research team took additional steps to protect Twitter user confidentiality by paraphrasing Tweet language rather than including word-for-word text to reduce reverse identification of Tweets (Ayers, Caputi, Nebeker, & Dredze, 2018). Moreover, all usernames, biographic information, and other potentially identifying information were removed from the dataset prior to coding.

In keeping with existing models of sentiment analysis, the research team utilized a combined inductive and deductive coding process to capture the nuance of online data (Bogen, Bleiweiss et al., 2019; Bogen, Bleiweiss, Leach, & Orchoswski, 2019; Bogen, Millman et al., 2018). A priori codes were taken directly from the WCQ, the diagnostic and statistical manual of mental disorders, fifth edition questionnaires to assess depression and PTSD (American Psychiatric Association, 2013), the five stages of grief (Kübler-Ross & Kessler, 2005), and measures of posttrauma resilience. Coding was conducted in NVivo by a three-person coding team, with two senior researchers in the field of traumatic psychology and clinical psychology advising on codebook creation, providing feedback on codebook development, and acting as consensus coders. Two coders spoke fluent Spanish and were familiar with slang and colloquial language in P.R., allowing the coding team to also analyze Spanish language Tweets. To be considered salient, themes had to emerge within 2% of the dataset (n = 12). The coding team established this cutoff after noting that critical emotional processes would otherwise have been excluded from the final dataset. The coding team met in real time, allowing for full thematic consensus.

The present study incorporated Geographic Information Systems (GIS) analysis to identify whether Twitter users’ IP addresses were classified as located in P.R., the mainland USA, or international sites. Inclusion of relevant GIS data allowed researchers to assess whether emergent themes varied as a function of user location, and how reactions to the death toll from Hurricane María differed for users in P.R. versus on the mainland USA. Geographic analyses were conducted in ARCGIS, and output maps were constructed utilizing ARCMap. Base map data was retrieved from Natural Earth and U.S. Census Cartographic Boundary Files, both of which provide public domain map datasets. GIS analysis was guided by a regional expert in spatial information science, geography, and data mapping.

Results

Thematic Content

Five major themes emerged from the 630 Tweets that utilized #4645Boricuas: discussion of phases of disaster responses (16%, n = 102), media coverage of the disaster (21%, n = 132), psychological process (55%, n = 345), politics (87%, n = 547), and resilience (9%, n = 59). An overview of these themes is presented in Table 1.

Table 1.

Themes among Tweets Using #4645Boricuas (N=630)

Theme	Subtheme	% (N)	Description	Example
Phases of Disasters	(Any)	16% (102)	Psychosocial reactions of survivors after a disaster.
	Disillusionment phase	15% (95)	Community and individuals realize the limits of the disaster assistance. Stress and discouragement start to arise due to the lack of help and support.	​“@realDonaldTrump tossing paper towels while saying Puerto Ricans should be proud of the assistance. Minimizing their situation and stating that Maria was not as bad as Katrina.” ​
Media Coverage	(Any)	21% (132)	Addressing commentary on media
	Commentary on the Media	32% (199)	General commentary on media	“All media coverage, tweets and politicians should talk about the genocide in Puerto Rico”
	Commendation	6% (37)	Positive comments about media	“After 8-months we are in the public eye thanks to the Harvard study. May this get us more help even if it’s too late. People are suffering because of Maria and because everyone forgot about them #PuertoRico #4645boricuas #4645GriefPR.”​
		4% (27)	Commendation for non-political leader	“@DavidBegnaud and @chefjoseandres are Puerto Ricans angels. Thank you for informing everyone and your hard work #4645boricuas”
	Criticism	11% (68)	Negative remarks on media	“@BrookeBCNN @johncolucci @iamsambee You should focus on #4645boricuas”
Psychological Processes	(Any)	55% (345)	Rationale provided for mental health or emotional disclosure
	Coping Narratives	8% (48)	Narratives about the death someone in the aftermath of Hurricane Maria	“My grandfather was diabetic and needed dialyses. Weeks after Maria, his treatments were reduced. He deteriorated quickly due to lack of electricity and limited food. He died 2 months after Maria.”
	Coping Strategies	30% (192)	(Any)
		20% (125)	Confrontation	“@realDonaldTrump, you should spend more time helping Puerto Ricans and less time twitting and sharing conspiracy theories. #4645Boricuas”​
		11% (72)	Seeking social support	“People are giving a voice to their families by sharing their names and stories. Don’t ignore them”
	Depression	7% (44)
		5% (33)	Feeling sad or depressed	“It’s difficult to be happy when the US, the country my grandfather served proudly doesn’t care about us”​
	Post-traumatic Stress Disorder	18% (114)	(Any)
		13% (80)	Disillusionment with authority	“#4645boricuas is hurtful… @CarmenYulinCruz said it best, we are dying @realDonaldTrump @ricardorossello You failed us, and you failed yourselves.”
		2% (14)	Distress when exposed to trauma reminders	“4645Boricuas is bringing back memories of the hurricane. It still hurts”
	Stages of grief	5% (29)	(Any)
		2% (14)	Acceptance	“There lay the shoes of those who died. #4645boricuas are no longer with us because of hurricane Maria and the government of PR” ​
Political Commentary	(Any)	87% (547)	Disclosing political views
	Advocacy	42% (266)	(General)	“#4645Boricuas is more than a hashtag. It is a reminder that our Puerto Rican siblings are dying and suffering. It is a reminder that the US government ignored them. Support them.” ​
		24% (152)	Corrective Information: #4546Boricuas, Puerto Rico, or Hurricane Maria	“#4645boricuas acknowledges the 4,645 lives lost in P.R. from September 20, 2017 through December 31, 2017, according to a Harvard study”
	Categories of Citizenship Norms	32% (202)	(Any)
		7% (47)	Participation: To be active in voluntary organizations, politics, and elections.	“People don’t understand. We are trying to keep Puerto Rico relevant. We need everyone to know about the death toll. Join instead of criticizing us”
		26% (165)	Solidarity: To support people who are worse off	“Neither us nor history will forget what happened in PR. We won’t forget the victims and victimizers.” ​
	Commentary on leadership	43% (274)
		2% (12)	Commendation: Positive commentary on political leadership	“@Carmen Yulin Thank you for your hard work and leadership!”
		42% (262)	Criticism: Negative commentary about political leaders	“#PuertoRico deserves better. It is unconceivable to die from hunger or lack of medical assistance. It’s a tragedy when something happens because of a hurricane, but negligence from the government is abhorrent. ​#4645Boricuas”​
		19% (117)	Criticism to Puerto Rican politicians	“The government doesn’t even know how to count dead people”
		4% (27)	Criticism to former First Lady Beatriz Rosselló	“@BeatrizRossello you care more about trivial things than the death toll. Puerto Ricans are dying and suffering under your and your husband’s watch. You are like Ivanka Trump, you have no feelings and no empathy towards your people. #4645Boricuas”
		10% 64)	Criticism to former Governor Ricardo Rosselló	“The governor of P.R. could have done something for P.R., but he refused. His lack of humanity killed us. He should be in jail instead of leading. He is a murderer.”
		24% (151)	Criticism to President Trump and GOP	“Trump and the majority-GOP Congress neglected #PuertoRico. The lack of assistance from the US government and the poor infrastructure of the island, were responsible, according to the Harvard study #4645boricuas”
	Perceived in group privilege	7% (43)	Perception that living in the mainland United States. opens more doors for non-Puerto Ricans in everyday life. Perception that non-Puerto Ricans have more privileges than Puerto Ricans.	“Puerto Rico is a commonwealth, not a state. Innocent Americans died.”​
	Reactions to communities impacted by natural disaster	4% (26)
		4% (26)	Respondents’ support for disaster assistance and their belief that federal government should provide assistance to victims of natural disasters.	“Puerto Rico is part of the US. They deserve equal treatment and assistance as other citizens after a disaster”​
Resilience		9% (58)	Community resilience indicators
	Resourcefulness	2% (12)	Identify problems, establish priorities, and mobilize resources; apply material and human resources to meet priorities and achieve goals.	“Join the petition to audit the debt in P.R., and death toll in P.R.. We will show the signatures to the United Nations in an open session.”
	Social	7% (47)	Social networks, social embeddedness, community values and cohesion	“Everyone, follow #4645Boricuas. People are sharing testimonies of the lives lost. Help give them a voice. Grateful for all who are following”

Phases of disasters

Tweets coded within the framework of DeWolf’s (2000) phases of disaster model primarily reflected the theme of disillusionment (n = 95). Specifically, Tweets coded as disillusionment reflected general frustration with the speed and responsiveness of the recovery efforts, and cited lack of disaster assistance, feelings of abandonment, and demands of assistance. For example, one Twitter user shared, “Trump and the GOP members of congress treated Puerto Rico after Hurricane María with lethal neglect. The Harvard study concluded the US government is to blame because they ignored the weak infrastructure of Puerto Rico.” Tweets commonly reflected the belief P.R. was neglected by the USA government, which these Twitter users felt provided inadequate aid.

Media Coverage of the Disaster

Tweets referring to media coverage represented 21% of the final sample. These Tweets mentioned news outlets, newspapers, tv stations, journalists, artists, or public figures to call attention to existing media or engage media personnel in coverage of relevant news. Users criticized the media’s inaction and commended them for their actions in bringing light to the situation. A small number of Tweets expressed gratitude and positive feedback to the media (6%, n = 37) whereas nearly twice as many Tweets (11%, n = 68) expressed frustration at the media for not sharing news about the hurricane or the death toll. For instance, one user tweeted, “@Username Can you cover the Puerto Rican lives lost, and share their stories? Can you tell me why the media has more interest in Trumps daily life than in P.R.? #4645Boricuas.” News outlets were also criticized for covering the personal lives of celebrities rather than documenting the tragedy and loss in P.R. These Tweets emphasized the importance of the media in keeping people informed and creating awareness for individuals outside the island.

Psychological Process

More than half the Tweets were classified as describing a psychological process (55%, n = 345). Data were first classified into coping narratives, operationalized as stories of how individuals coped with the disaster. Tweets were further categorized into distinct measures of coping, including confrontation and seeking social support. Data were also coded to reflect specific psychological symptomatology, including depression, PTSD, and stages of grief.

Coping Narratives.

A small sample of Twitter users utilized #4645Boricuas to share stories of people they knew who died during Hurricane María or in the aftermath. Surprisingly, these Tweets made up only a small sample of the dataset (8%, n = 48). One user shared, “#4645Boricuas Juliana’s father is one of the #4645Boricuas. He died of complications after he fell in his home at night. Lack of electricity was the cause.” Tweets provided examples of how the lack of assistance, water, and electricity after Hurricane María was fatal.

Coping Strategies.

Overall, 30% (n = 192) Tweets mentioned coping strategies. Of coping strategies outlined in the WCQ, Twitter users primarily expressed confrontation (20%, n = 125), and seeking social support (11%, n = 72). Tweets that cited confrontation included comments that expressed anger, requested assistance, or reflected the hope that people in power – either the media or government – would act to help. Tweets also reflected upon the perceived lack of help and empathy. For instance, one Tweet stated, “Waiting for @ricardorossello to understand that having the flag at half-mast is not in vain #4645Boricuas #PuertoRico.” Tweets categorized as social support-seeking included comments that asked for emotional support, especially sharing the news about the death toll via Twitter. For example, “To help tell the stories of their loved ones or show your support for #PuertoRico use #4645Boricuas to spread the word!” Twitter users also expressed desire for sympathy.

Depression.

A subset of Tweets (7%, n = 44) mentioned symptoms of depression, highlighting the psychological distress individuals experienced following the news of the death toll. Most users expressed feelings of sadness, for example, “How are you? (in my head) “I cry every day for myself, my country, my people, its future. My heart aches and there are days I don’t want to get out of bed’—Great! #PuertoRico #4645Boricuas #PR.” Tweets specifically expressed sadness about the lack of assistance.

Post-traumatic stress disorder.

Tweets also described symptoms of PTSD (18%, n = 114). The news of the death toll may have been triggering for individuals who were either directly or indirectly impacted by the hurricane. Researchers used the Clinician Administered PTSD Scale (Blake et al., 1995) to assess for thematic mentions of PTSD. The two salient symptom sub-categories emerged, including disillusionment with authority (13%, n = 80) and distress when exposed to trauma reminders (2%, n = 14). Tweets reflecting disillusionment with authority included frustration regarding how the local and federal governments and politicians managed hurricane relief, and the news of the death toll. For example, “#4645boricuas #4645Americans My heart breaks for them. People died on their way to hospitals or lack of food, electricity, and supplies. The Trump administration hasn’t helped and lies about the death toll.” Additionally, a small subset of Tweets reflected experiencing distress when exposed to reminders about Hurricane María. Tweets referenced the impact that the news was having on them, with one user stating, “These stories break my heart. The government won’t recognize so many deaths. #4645boricuas.” Tweets reflected the ongoing impact of the hurricane, and the continued feelings of sadness.

Stages of Grief.

Several Tweets also reflected stages of grief (5%, n = 29), as described by Kubler-Ross (1973). Acceptance was the most salient category of Tweets expressing grief (2%, n = 14). Notably, 10 out of the 14 Tweets were written in Spanish. This could indicate a higher likelihood that Puerto Ricans wrote those Tweets. One Tweet stated, “The #4645Boricuas who died are speaking louder and clearer than any of us. We have to step aside so that they can ask for justice.”

Political Commentary

The most common theme within the sample was political commentary (87%, n = 547). This theme included several subthemes: (a) advocacy, (b) categories of citizenship norms, (c) commentary of leadership, (d) perceived ingroup privilege, and (e) reactions to communities impacted by natural disasters.

Advocacy.

Approximately 42% (n = 266) of the total Tweets advocated on behalf of P.R. and Puerto Ricans who suffered during Hurricane María and its consequences. Tweets classified as advocacy urged others to help Puerto Ricans, or condemned negative actions or the lack of action from the government or the media. For example, “More coverage for the lives lost in Puerto Rico and their stories! The media pays more attention to Trump’s daily life than to Puerto Ricans, and how to prevent future deaths. #4645Boricuas.” Additionally, 24% (n = 152) of Tweets characterized as advocacy included corrective information, such as information about Hurricane María, P.R., or Puerto Rican citizenship.

Citizenship norms.

A significant proportion of the dataset included themes discussing citizenship norms (32%, n = 202). Out of the four categories of citizenship norms outlined by Dalton (1998), only two were salient: (a) participation (7%, n = 47) and (b) solidarity (26%, n = 165). Tweets were coded as mentioning participation if they included comments on engaging in voluntary organizations or politics, or mentioned ways that others could help P.R. For instance, “Many of us have the privilege of forgetting, but we can also have the privilege of sharing, donating, and speaking up. Do it, but don’t let yourself forget! #4645boricuas” Additionally, Twitter users expressed their solidarity with P.R. through providing comments of support for P.R., Puerto Ricans, or as Dalton (1998) classifies it, expressing support for people in more dire circumstances. For example, “#4645Boricuas are American citizens, they should receive the same assistance and respect other American citizens receive after a disaster.”

Leadership.

Many Tweets coded as political commentary referenced leadership (43%, n = 274). Twitter users expressed their frustration towards political figures and criticized or commended politicians. Most of the comments on leadership were critiques (42%, n = 262) rather than praise. For instance, “#4645Boricuas represents the denial from the government on the true death toll instead of the reported 64 for the last year. This number supports that the government ignored the deaths.” Tweets were frequently directed at the USA government, President Trump, and the GOP (23%, n = 151). These Tweets criticized their approach and lack of empathy, and in some cases discrimination towards Puerto Ricans. One user shared, “@PressSec @POTUS The response and lack of compassion from @realDonaldTrump to #PuertoRico was historic. He’s the only president to display disrespect, discrimination, incompetence & disregard to people harshly hit by a natural disaster. #4645Boricuas.” The Puerto Rican government was also criticized by Twitter users (19%, n = 117). Criticism was commonly directed to former Governor Ricardo Rosselló, (10%, n = 64), and former First Lady Beatriz Rosselló (4%, n = 27). For instance, “Ricardo Rosselló debated the reports of death tolls higher than 64, and his administration wouldn’t release official information. The first lady joked about the idea of a memorial. Today they both visited the memorial. #4645boricuas.”

A very small category of users wrote Tweets with positive comments and thanked political leaders for their work and response to Hurricane María (2%, n = 12). For example, “Thank you @KamalaHarris for using your voice to share the news of #PuertoRico and keep to the memory of our #4645Boricuas alive. We appreciate your compassion!” It is worth noting that no Tweets included a positive commentary on President Trump and the Republican Party, and only one commendation referenced former Governor Rosselló.

Perceived ingroup privilege.

The conceptualization of ingroup privilege was adapted from the framework proposed by from Leach, Iyer and Pederson (2006). In this study, these concepts were adapted to reflect the relationship between P.R. and the USA, including (a) “non-Puerto Ricans have certain privileges that Puerto Ricans do not have in this society” and (b) “I feel that living in the mainland United States opens more doors for non-Puerto Ricans during their everyday lives.” Because of the similarities of these two items, and because Tweets referenced both multiple times, they were merged into one category (7%, n = 43). One user shared, “The U.S. has failed and continues to fail us. We will continue to be second class citizens. We’ve been colonized, disregarded, and neglected. #4645Boricuas.” Tweets within this category included comments regarding injustices to Puerto Ricans, discrimination, or Puerto Ricans as second-class citizens. Twitter users also shared their feelings about the status of P.R. as a commonwealth, and the different ways that the government responds to victim of natural disasters in the mainland USA vs. the island of P.R.

Reactions to communities impacted by natural disasters.

Tweets reflected items in the General Support for Disaster Assistance Questionnaire (Skitka, 1999); in particular, the belief that “it is the appropriate role of the government to provide assistance to victims of natural disasters”. Approximately 4% (n = 26) Tweets reflected this belief. For Tweets to be classified under this category, Twitter users had to mention the USA or Puerto Rican government, or Puerto Ricans as USA citizens and advocate for disaster assistance. For example, “33% of the #4645boricuas that died because of #HurricaneMaría, died due to lack of access to medical care. The suffering in #PuertoRico is not over. The Trump administration needs to do its job and stop playing with the lives of these Americans.” Twitter users indicated that Puerto Ricans deserved more robust disaster relief.

Resilience

The final theme identified from the dataset was resilience (9%, n = 58), reflected in the Dimensions of Resilience measure (Bruneau et al., 2003). Tweets highlighted how individuals assessed, responded to, and strategized to cope with the news of the death toll. Resilience for physical and social systems include: (a) robustness, (b) redundancy, (c) resourcefulness and (d) rapidity (Bruneau et al., 2003). Community Resilience Indicators (Cutter et al., 2008) were also utilized to guide the classification. Resourcefulness and Social Resiliency emerged as salient sub-classifications and are described below:

Resourcefulness.

The most salient dimension of resilience was resourcefulness (2%, n = 12), characterized by identifying problems, organizing and preparing resources, and using available resources to meet the objectives (Bruneau et al., 2003). One user shared, “With more than 100 signatures to #AuditThePRDeaths #AuditTheDebt for #PuertoRico #4645Boricuas. In solidarity, we will join the #PRParade and go to the @UN! SIGN AND SHARE!” Tweets included commentaries of available resources and invited other users to do the same.

Social Resiliency.

Social resiliency was cited by approximately 7% (n = 47) of Tweets within our dataset. Tweets mentioned social networks, community values, and cohesion (Cutter et al., 2008). For example, “You help us by caring and reminding others! Thank you! #4645Boricuas.” Tweets categorized as reflecting social resilience commonly expressed gratitude toward their communities, media, and politicians who provided assistance.

Analysis of 5 top Tweets

Overall, the most popular 5 Tweets in the original dataset of 33,549 Tweets reflected primary themes of political commentary (100%, n = 5), mental health (60%, n = 3), and media (20%, n = 1). Results from the top Tweets are summarized in Table 2. In total, the top 5 Tweets were retweeted over 23,000 times and had the potential to reach over 2.5 million Twitter users (the total number of followers for the individuals who shared the top 5 Tweets). It is notable that all of the influencers who shared the most popular Tweets discussed political views regarding the estimated deaths of Hurricane María. Individuals with social power (i.e., celebrities) may feel compelled to use their power in an influential way following a natural disaster.

Table 2.

Top Tweets

Tweet Text	Number of Retweets	Number of Followers
1. RT @andreagonram: Since the news cycle and Twitter dot com apparently already forgot about the 5,000+ people who died in Puerto Rico after Hurricane Maria, I’m going to spotlight the stories of our dead. #4645boricuas	13,560	5,483
2. RT @Lin_Manuel: More Americans dead than September 11th. Our president threw paper towels and moved on to the next outrage. #4645Boricuas	6,570	2,361,878
3. RT @SarahMolinari: Hundreds of shoes placed in front of the Capitolio to symbolize and protest the #4645boricuas who lost their lives due to post-Maria shameful neglect, interruption and denial. A powerful public mourning, honoring and remembering of those who were never counted.	1,201	2,96
4. RT @yarimarbonilla: In the wake of a newly released study, Puerto Ricans are taking to social media (and the streets) to call attention to the 4,645 reported deaths as a result of hurricane Maria. Follow #4645boricuas to read their stories. Join protests on June 1 to demand that the dead be counted.	1,002	3,289
5. RT @AbbyMartin: A shocking 4,645 people died in Puerto Rico after Hurricane Maria, more than Katrina & 9/11 combined. 70x higher than what the govt told us. @rosaclemente exposes the criminal negligence that killed them in “Puerto Rico Rising” #4645boricuas #PRonthemap	773	213,483
Total	23,106	2,564,374

Analysis based on Tweet location

GIS analyses (Table 3) revealed important distinctions of thematic frequency, based on whether Tweet originated from the island of P.R. or the mainland USA. Of Tweets discussing Disaster Phases, 18% (n = 30) originated from P.R., 34% (n = 55) originated from the mainland USA and 48% (n = 79) originated from international countries (operationalized as any location outside the mainland USA, Hawaii, Alaska, and P.R.). Notably, of Tweets discussing Media Coverage, 22% (n = 29) originated from P.R., 65% (n = 84) originated from the mainland USA, and 13% (n = 17) originated from International countries, indicating that users in the mainland USA were most likely to discuss Media Coverage perhaps as a way to “signal boost” links, narratives, and other messages originating from the island (see Figures 1–5).

Table 3.

Results Per Theme and Geographic Origin

Theme	Total (N = 630) N (% overall dataset)	Puerto Rico n (% within theme)	Mainland US n (% within theme)	International n (% within theme)
Disaster Phases	164 (26%)	30 (18%)	55 (34%)	79 (48%)
Media Coverage	130 (20%)	29 (22%)	84 (65%)	17 (13%)
Psychological Processes	336 (53%)	118 (35%)	178 (53%)	40 (12%)
Political Commentary	533 (84%)	190 (36%)	276 (52%)	67 (12%)
Resilience	55 (9%)	22 (40%)	29 (53%)	4 (7%)

Figure 1.

Map of origin of Tweets discussing Disaster Phases

Figure 5.

Map of origin of Tweets discussing Resilience

Furthermore, 35% of Tweets (n = 118) discussing psychological processes originated from P.R., whereas 53% (n = 178) originated from the USA, and only 12% (n = 40) originated from international sources. Though most Tweets discussing the theme of Psychological Processes originated from the mainland USA, the proportion from P.R. increased drastically compared to themes of Phases of Disasters and Media Coverage, indicating that Twitter users in P.R. were more concerned with discussing the psychological outcomes of the disaster and news regarding the death toll. Similarly, 36% (n = 190) of Tweets discussing the theme of Political Commentary originated from P.R., while 52% (n = 276) originated from the mainland USA, and only 12% (n = 67) originated from International sources. Once again, Twitter users from P.R. engaged significantly with online discourse surrounding politics, including condemnation of political figures perceived as ineffective in the aftermath of the hurricane, as well as praise for local leaders. Finally, of Tweets discussing the theme of Resilience, 40% (n = 22) originated from the island of P.R., 53% (n = 29) originated from the mainland USA, and merely 7% (n = 4) originated from international sources.

GIS results indicated that Twitter users around the globe were discussing the death toll from Hurricane María. However, despite the relatively small number of citizens of P.R. (3.165 million) versus the population of the mainland USA (306.675 million, including Alaska and Hawaii) and the global population (7.7 billion), a significant plurality of Tweets in the dataset nevertheless originated from P.R. It is important to note that the population of the mainland USA is approximately one hundred times as large as the population of P.R., and the global population is approximately 2,432 times as large as the population of P.R. Therefore, Tweet proportions overall were heavily skewed toward originating from P.R., across theme groups. Overall, Tweets originating from P.R. were significantly more likely to discuss theme groups of Psychological Process, Politics, and Resilience compared to either Disaster Phases or Media.

Discussion

The present study examines Tweets with the hashtag #4645Boricuas to examine public discourse surrounding the announcement of the death toll resulting from Hurricane María. Consistent with previous research examining public sentiment surrounding natural disasters on Twitter (Gruebner et al., 2017), the coding of Tweets revealed that users who tweeted months after the hurricane reflected language suggestive of the disillusionment phase. Users also tweeted about mental health outcomes as a result of the hurricane, describing symptoms of depression and PTSD, as well as evidencing a range of coping strategies. Not surprisingly, researchers observed that both media coverage and political commentary were commonly described by Twitter users as important to the collective response to the hurricane. There were also several salient subthemes that emerged within these major themes.

Past research has characterized the phasic nature of community and individual level coping with disasters, coding included pre-disaster, impact, disillusionment, and reconstruction phases (deWolfe, 2000). Tweets categorized to be in the disillusionment phase frequently compared the response to Hurricane María relative to other natural disasters, noting inadequacies in the resources provided following the hurricane. This pattern is not surprising given that sampling of Tweets occurred about nine months post-trauma, as researchers have proposed that the disillusionment phase tends to range from 6-months to one- or two-years post-discharge (Ollendick & Hoffmann, 1982). As noted by DeWolfe (2000), the disillusionment phase following a disaster may include triggering events, which extend disillusionment among survivors. The announcement of the 4,645 estimated deaths could have been a trigger. The likelihood of reflecting on disillusionment is also consistent with historical events that coincided with the present sampling period. Specifically, #4645Boricuas as a reference to efforts to document and remember lives lost through Hurricane María. Findings from this study continue to point to the importance of providing continued resources and support to disaster-affected communities well after the honeymoon phase.

The disillusionment phase has also been associated with increased community and individual level struggles with mental health and psychological wellbeing. A large proportion of Tweets (18%) included references to posttraumatic stress symptomatology, including diagnostic symptoms of PTSD, and features associated with PTSD. Tweets reflecting disillusionment with authority were common, perhaps in part due to the timing (9 months postdisaster) as well as the political and historical events at the time of the sampling. Importantly, when communities experience changes in overall belief/faith in authorities and institutions, it can interfere with the transition of communities in rebuilding and moving forward, and can impede individual-level healing. Although early assessments of PTSD included disillusionment with authority as an associated feature, this concern was described to be population specific and it is not included in current gold standard clinical interviews (see Weathers, Keane, & Davidson, 2001).

Tweets also reflected the psychological impact of the disaster. There is a rich literature suggesting that PTSD and depressed mood frequently co-occur and have shared risk factors (Campbell et al., 2007; Goldstein et al., 2016; Spinhoven, Penninx, van Hemert, de Rooij, & Elzinga, 2014; Wisco et al., 2016). In the present sample, there were somewhat fewer Tweets related to depressive symptomatology, with the most widely endorsed depressive symptoms tending to be those that are also a part of PTSD, such as feeling sad, distressed, or emotionally numb. Given that this is less commonly described in the current sampling of Tweets, it is possible that among users choosing to Tweet, these feelings of sadness or depression were secondary to other, more frequently highlighted, feelings such as posttraumatic stress and disillusionment. It is possible that, over time, the prevalence of depressive feelings may change as a function of transition to later phases post-disaster.

A number of Tweets described the deaths of family and friends that users attributed to Hurricane María and post-disaster inadequacies of healthcare and resource provision. A wealth of research points to the negative effects of traumatic bereavement, with trauma-related death of friends and loved ones complicating and often slowing healing and post-trauma intervention (Barlé Wortman, & Latack, 2017). Tweets related to deaths ranged from reports of death secondary to the hurricane that happened during the acute phase of the disaster as well as deaths that happened later but were attributable in part to the secondary effects of the disaster on the community infrastructure; accordingly, whereas many references were to individuals who had died in more recent months, a majority referenced loss of loved ones that happened more than 6 months prior. Consistent with the passage of time since these losses, very few Tweets involved the earliest stages of grief, such as denial. The most commonly observed stage of grief was acceptance, with the majority of these Tweets written in Spanish. The present sampling period was prompted by a research report that pointed to mortality estimates well beyond what had been previously suggested in public forums. Accordingly, it is not surprising that some Tweets sought to tie individual stories, losses, and people to the number presented in the research. Indeed, several Tweets noting deaths of loved ones included statements with direct pleas to give “voice” to both those who have died and to the bereaved families who want to be heard. These tweets also allowed individuals to connect and to ‘express their emotions, concerns, well wishes, and to memorialize victims’ of Hurricane María, which is another disaster social media use in Houston et al. (2015) framework.

Consistent with both the timing of the sampling period as well as historical events occurring at the time of the sampling, Tweets were most likely to focus on confrontation as a coping approach and were less likely to focus on positive reappraisal. Social support was also observed to be a prevalent coping strategy; notably, however, interpersonally focused Tweets sampled during this time frequently included overtones of disillusionment. Prior work in disaster exposed samples has supported the buffering model of social support, which proposes that the negative impact of stress may be mitigated by social relationships (Aneshensel & Stone, 1982; Gaston et al., 2016). Indeed, categorization of Tweets judged to fall in the resilience category were most likely to focus on social networks, highlighting community cohesion and promoting connectedness among the community.

Consistent with the utilization of social media as an outlet for hashtag activism (Yang, 2016), the most prevalent type of Tweet consisted of the expression of political views. This consisted of commentary explicitly critiquing or commending specific politicians or governmental action, as well as a call to action to elect governmental officials who will promote stronger support for P.R. in response to disasters. Tweets criticizing and blaming the USA and P.R. government, political leaders, and the political status of P.R. reflect the disaster social media use developed by Houston et al. (2015) in the post-events disaster phase ‘discuss socio-political causes and responsibility for events’. Individuals also commented on the need to support others influenced by tragedy. It was notable that a small number of Tweets expressed concern that Puerto Ricans living on the mainland were afforded greater privilege than those living on the island. Indeed, there are significant public health crises in P.R. that predate the hurricane (Alegria et al., 2001; Oropesa & Landale, 2000).

The present study also highlights the importance of recognizing resilience as a vital component of the disaster response. It is now well documented that the assumption of an automatic link between the experience of trauma and psychological symptomatology obscures the opportunity for post-traumatic growth among survivors of natural disasters (Patton, Smith, & Violanti, 2000). Relational networks are important catalysts for healing and posttraumatic growth (Walsh, 2007), and Twitter may serve as an outlet for individuals to share common experiences, connect with others who share similar concerns, garner social support, and reconnect with community members (Houston et al., 2015). Puerto Ricans living on the island might not have been the only ones trying to reconnect with their community, people who experienced Hurricane María and later moved to the mainland USA, could have used #4645Boricuas to reconnect with their community members and other people who shared their experiences. Notably, there was little if any dialogue within #4645Boricuas that would be labeled as harassing, derogatory, or “trolling” – a term utilized to describe antagonistic behavior in online spaces. Thus, we can surmise that this online space was generally supportive for the individuals who shared their commentary by using #4645Boricuas. Further research is warranted to understand whether participating in this online dialogue was a positive experience.

Although it is unknown whether individuals tweeting from the mainland USA identified as Puerto Rican, it was notable that the #4645Boricuas was utilized by individuals tweeting from both locations, as well as across the globe. Given the relative population sizes of P.R. and the mainland USA (the population of the mainland USA is approximately one hundred times greater than the population of the island), it is particularly important that a large plurality of Tweets originated from P.R. Findings indicate that people in P.R., given relative geographic closeness to the hurricane and immediacy of needs, were proportionally more likely (when taking into account population sizes) to discuss the death toll from Hurricane María on Twitter. Similarly, whereas international users were most likely to discuss Disaster Phases, individuals tweeting from P.R. were focused primarily on Resilience and Politics, while Twitter users in the USA most commonly shared Tweets concerning the Media theme (i.e. links to relevant stories).

Special reports by the World Bank on the use of geomapping in disaster relief efforts assert that visualizing locations in need of specific resources allows aid organizations to better monitor their impact, improve aid effectiveness and coordination, and enhance transparency of relief efforts (Gigler, Tanner, & Kiess, 2011). The content of Twitter users in P.R.’s discussions of psychological process and resilience indicate that people on the island may be in specific need of psychological resources and information regarding trauma coping. It may be useful for clinicians and resource centers to provide psychological support to people in P.R. via the internet, as the island waits for infrastructure to rebuild.

Limitations

The present study has limitations, and findings should be understood in consideration of these restrictions. First, the Twitter character limit results in brevity which may artificially influence the content or tone of user conversations. In contrast, naturalistically observed oral conversation or even online messaging platforms that do not limit the character number may provide different information. Second, the research staff analyzed Tweets exclusively using #4645Boricuas and each Tweet was analyzed as a unit. As such, it is feasible that relevant Tweets not including the hashtag were not captured and analyzed. Therefore, results are not generalizable to all Twitter discourse regarding Hurricane María or the news of the death toll. Additionally, the research team did not code for discussion between users within the sample (i.e., a “back and forth” dialogue on Twitter), meaning that analysis possibly lacked conversational nuance. Third, Tweets without latitude and longitude data to permit geomapping and Tweets shared outside of the time point of data collection were excluded from the analysis reducing our dataset and thus not including Tweets of people living in P.R. who expressed themselves on Twitter. It is also important to note that not everyone uses this social platform or has access to Twitter. Finally, one important thing to consider when we talk about access is that data for this study was collected seven months after Hurricane María. At that time, many people living in P.R. were still having problems with internet connectivity and still did not have the electrical service restored. Therefore, the results are not generalizable.

Clinical Implications

Despite of these limitations, the findings from the current study have important clinical implications. Our findings suggest that the online discussion about the estimated death toll of Hurricane María promoted the disclosure of emotions and reactions after trauma. There is literature supporting the health benefits of writing about traumatic events (Smyth, True, & Souto, 2001), and how it can help people with grief (Furnes & Dysvik, 2010). Part of the benefits shown in our findings was that sharing online disclosures elevated community sense of belongingness, social resilience, and advocacy. Also, it was useful for people who were hoping to receive immediate emotional support (Lachmar, Witterborn, Bogen & McCauley, 2017; Shepherd, Sanders, Doyle, & Shaw, 2015). Further research is needed to evaluate the impact of participating in an online forum that includes trauma narratives and vivid accounts of human loss.

The present study also highlights the unity of GIS as a resource to identify communities in most need of medical or psychological assistance based in their online disclosures. These findings provide value information for mental health professionals and first responders can use when working in the aftermath of a disaster or other type of emergency. The use of GIS can also assist in ensuring that online interventions, apps, and educational resources are disseminated to those communities who are most in need. For example, it is feasible that Twitter data could be utilized in the future to identify communities where public sentiment is particularly problematic, so that resources could be targeted towards these communities in need. The research findings also point to the importance to understand the socio-political situation of P.R. and how it impacts their psychological processes (Capielo Rosario, Abreu & Gonzalez, 2018), their experiences, and the recovery after a natural disaster. Training of clinicians working with Puerto Ricans and other Latinx populations should consider the evaluation of the socio-political conditions and how it has affected the individual.

Conclusion

As natural disasters continue to impact communities across the globe, it is essential to understand how communities cope with this unique form of widespread trauma. Social media represents an opportunity for individuals to share their experiences, and express solidarity. The present study highlights ways in which individuals in P.R., the mainland USA, and internationally are connected using #4645Boricuas to discuss disillusionment with the disaster response, raise awareness and call media attention to the death toll, process the psychological impact of the experience and share coping skills, express political views, and show their resiliency. The findings may be useful in understanding how to support other communities in coping with disasters in the future.

Figure 2.

Map of origin of Tweets discussing Media Coverage

Figure 3.

Map of origin of Tweets discussing Psychological Process

Figure 4.

Map of origin of Tweets discussing Politics