Abstract
People who are affected by cancer can benefit greatly from social support and digital social networks, though our understanding of online support is primarily founded in dominant platforms like Facebook. In addition, while previous scholarship indicates that social support is available online, little research has examined predictors of support provision. A content analysis was performed to examine the relationship between narrative features in Imgur posts and social support in comments. Imgur (Imgur.com) is a social media site and image-hosting platform, amassing over 250 million monthly visitors. Six post features were hypothesized to predict support, including explanations of the diagnosis experience, evidence of agentive problem solving, indications of positive reappraisal, pleads for the audience to get a checkup, references to mortality, and inclusion of humor. The results of this study indicate a relationship between narrative construction and social support, finding that the inclusion of narrative features in cancer-related posts influenced the provision of support in comments. Findings of this study could have implications for a multitude of stakeholders interested in social support provision, including healthcare professionals and researchers interested in the use of social media platforms for support, and organizations interested in designing supportive online platforms for individuals coping with cancer.
Online communities can be a source of support for individuals coping with stressful situations, including illnesses like cancer (e.g., Rains & Young, 2009; Wright, 2002), heart disease (Lindsay et al., 2009; Tak & McCubbin, 2002), and depression (Evans et al., 2012; Rains et al., 2016). Online support networks are especially beneficial to individuals with limited mobility, weak or unsupportive offline support networks, and/or rare illnesses (Rains & Wright, 2016; Sposito et al., 2015). The proliferation of digital technology and connectivity has expanded the opportunities for online support networks to develop, and research has shown these can vary greatly (Hayes et al., 2016; Meng et al., 2017; Rains et al., 2015). While early digital research explored more professionally moderated support forums (e.g., White & Dorman, 2001; Wright, 2002), research has since expanded to include social networking sites—most prominently Facebook (Buehler et al., 2019; Gage-Bouchard et al., 2017; High et al., 2014; Vitak & Ellison, 2013). However, recent reviews of the support literature have called for research on additional digital platforms (Meng et al., 2017; Rains et al., 2015), allowing for more comprehensive assessments of the functions of support in social media and increased generalizability of research inquiries.
In addition, while previous scholarship has provided descriptive analyses of online support availability and the classification of support types, comparatively little research has examined predictors of support in networked communities on social media (for exceptions see Davis et al., 2015; Hale et al., 2018; McLaughlin et al., 2012). In their review of support literature from 2004 to 2015, Meng et al. (2017) found that only 6.8% of reviewed studies evaluated antecedents to support provision, including characteristics of support seekers (e.g., support availability, family interaction, and narcissism) and message features that elicited support (e.g., self-disclosure and the inclusion of family and prayer in posts). Therefore, while the availability of support through digital platforms has been established in previous scholarship, more work is needed to determine the factors that engender online support.
In response to calls for further research evaluating antecedents to support provision and across social media platforms, a content analysis was performed of 117 cancer-related Imgur posts and 2511 corresponding comments, examining the relationship between post features and social support in comments. The Imgur platform falls in a genre of social networks defined by a culture of anonymity (which also includes sites like Reddit and Tumblr), in which the expectation of semi-anonymity influences engagement and discourse (e.g., Brown et al., 2018; Massanari, 2017). Moreover, the voting system utilized by the platform (a bidirectional setup using upvotes and downvotes) allows users to influence how content is displayed in the site (similar to Reddit). Accordingly, the architectural and cultural dissimilarities between the Imgur platform and Facebook, which has received the bulk of previous scholarly attention, uniquely positions Imgur for comparison with previous social support research. Therefore, this work contributes to social support research by examining the relationships between post features and the provision of support in comments, helping us better understand online support exchange. Findings of this study have implications for stakeholders interested in social support provision, including researchers, healthcare professionals, and organizations interested in designing supportive online platforms for individuals coping with cancer.
Social Support and Social Media
According to Berkman (1984), social support is defined as “the emotional, instrumental, and financial aid that is obtained from one’s social network” (p. 415). For individuals dealing with health issues like cancer, the availability and reception of social support can have important implications for psychological and physical health outcomes (Berkman, 1984; House et al., 1988; Uchino et al., 1996). Prior research suggests that support is a significant factor in facilitating successful coping processes (DeLongis & Holtzman, 2005), seems to mediate stress and well-being (Cohen & Wills, 1985; Dunkel-Schetter et al., 1987), and helps individuals overcome challenges (Hale et al., 1997). In addition, online social support networks attenuate spatial boundaries that often accompany long-term illness (Sposito et al., 2015). Compared to other health communities, the cancer community is one of the most supported and supportive (Dakof & Taylor, 1990; Davison et al., 2000; Walther & Boyd, 2002; Wright & Bell, 2003), with established support resources for cancer patients to connect with medical professionals and survivors.
Despite a growing variety of social networks online, previous online social support research has predominantly focused on Facebook (e.g., Buehler et al., 2019; Gage-Bouchard et al., 2017; High et al., 2014; Vitak & Ellison, 2013), which makes sense considering the popularity of the platform. However, Meng and colleagues (2017, p. 48) argue that the preeminence of Facebook in support research may “privilege a particular group of people and cultural practices” hindering our ability to infer generalizable practices of digital support exchange. Other social media platforms have distinct features that affect their site culture, such as the levels of anonymity/pseudonymity they afford, character restrictions imposed upon comments, and the ranking of comments according to user feedback (e.g., votes), potentially resulting in a different sense of community and group identity (Barnes, 2018; Reagle, 2015). This study expands support research through an examination of Imgur, which is characterized by a culture of anonymity, humor, and image-based communication (Hale, 2017; Mikal et al., 2014). Considering the sensitive nature of life-threatening illness, individuals could at times prefer anonymous platforms like Imgur over identifiable sites like Facebook for disclosing fears and seeking support (Walther & Boyd, 2002).
Imgur (Imgur.com) is a social media site and image-hosting platform, amassing over 250 million monthly visitors (Imgur, 2019b). Carr and Hayes (2015) define social media as “Internet-based channels that allow users to opportunistically interact and selectively self-present, either in real-time or asynchronously, with both broad and narrow audiences who derive value from user-generated content and the perception of interaction with others” (p. 50). Conforming to this definition, Imgurians (Imgur users) interact by (1) sharing visual and textual user-generated content through posts (at least one image is required per post), (2) responding to posted content using short (140 characters) comments, and (3) voting on posts and comments using “upvotes” and “downvotes,” which signify like and dislike, respectively. Although users remain principally anonymous, the Imgur network predominantly consists of Millennial men (Imgur, 2019a), similar to Reddit (Statista, 2019), which differs from networks like Facebook where userbases are demographically heterogeneous across gender, age, and income (Weareflint, 2018).
The pseudonymity of Imgur could have implications for social support provision. To create an Imgur account, users disclose and verify limited personal information (e.g., their phone number, email, and/or connection to another social network), though this information is not accessible to other users. Subsequent user interactions are pseudonymous (similar to Reddit), though posting and commenting history is preserved by the platform (unless deleted by the user). Architecturally, Imgur centralizes interaction within a “front page” in which Imgurians view the same content and interact with one another through popular community-selected posts. This differs from sites like Facebook which utilize a user-centered design (i.e., the user is positioned as the central node of their social space), YouTube, which uses a channel-based framework (i.e., users “subscribe” to channels), or Reddit, which employs a community-of-interest design (i.e., subreddits). With content as the central node rather than the direct connection with others, an individual channel, or community-of-interest, support provision becomes contingent on post details and cultural norms, as opposed to user identity and corresponding social norms. Because of these differences in network size, demographic homogeneity, site affordances, and architecture, Imgur is well-positioned for comparison with previous examinations of platforms like Facebook, YouTube (e.g., Hale et al., 2018; Liu et al., 2013), and Reddit (e.g., De Choudhury & De, 2014). In particular, while previous work has suggested that social media platforms are useful for eliciting social support, an analysis of support provision by Imgur users should provide a unique contextual comparison. Thus, the first research question is proposed to assess the supportiveness of Imgur discourse about cancer-related posts:
RQ1. How frequently will social support emerge in comments submitted in response to cancer-related Imgur posts?
Types of Social Support
In addition to addressing supportive communication generally, previous scholarship has identified a variety of support types, and several typologies of social support have been extended (Cutrona & Russell, 1990; Cutrona & Suhr, 1992; Neuling & Winefield, 1988; Rains et al., 2015). This study evaluates four support types outlined by the Multi-Dimensional Support Scale (MDSS) (Neuling & Winefield, 1988)—reassuring, empathic, informational, and tangible support. Reassuring support provides confidence through words of affirmation or hope (e.g., “You can do this!”). Empathic support is provided to others in environments of acceptance or love to facilitate an understanding of the individual’s issues (e.g., “I understand where you are coming from” or “I see how you are feeling”). Informational support is provided to organize thoughts and provide appraisal for the individual, including advice (e.g., “Perhaps consider a diet change”). Finally, tangible support involves direct aid through financial or physical assistance (e.g., “Can help cover your healthcare costs?”). The relative prominence of these support types fluctuates in online communities, with research finding various levels of empathic or informational support (Blank et al., 2010; Buis & Whitten, 2011; Ginossar, 2008; Love et al., 2012), informational support (Ginossar, 2008), and reassuring support (Hale et al., 2018). Accordingly, the following research question is extended:
RQ2. How frequently will reassuring, empathic, informational, and tangible support emerge in Imgur comments submitted in response to cancer-related Imgur posts?
Narrative
Previous research suggests that narrative construction can be a beneficial coping strategy for individuals dealing with chronic illness (Kreuter et al., 2007), as the purposeful construction of a narrative allows patients to assign causality for their condition, determine linearity of the illness process, and produce a framework through which their situation can be understood (Kellas, 2016; Kellas & Manusov, 2003; Ziebland & Wyke, 2012). For cancer patients in particular, creating a narrative can help regain a sense of agency or control over the situation (Bishop & Yardley, 2004; Chou et al., 2011; Midtgaard et al., 2007) encourages a sense of self-reliance (Midtgaard et al., 2007) and allows patients to “renegotiate agency within the medical system that positions doctors as experts and cancer patients as merely recipients of treatment and information” (Hale et al., 2018, p. 576). Beyond constructing a cancer narrative, sharing with others may be especially beneficial, as audiences respond by providing appropriate forms of support, facilitating psychological adjustment and promoting recovery for the patient (Berkman, 1984; House et al., 1988; Neuling & Winefield, 1988; Uchino et al., 1996). Moreover, the response received from readers or listeners is partially determined by the narrative structure employed by the patient, as this helps the audience “make sense of [their] own situation by suggesting a practical and emotional frame for [their] response” (Ziebland & Wyke, 2012, p. 236).
Previous scholars have identified four cancer narrative features connected with support provision: explanations of the diagnosis experience, agentive problem solving, positive reappraisal, and pleads for audience checkup (e.g., Chou et al., 2011; Dunkel-Schetter et al., 1987; Hale et al., 2018; Liu et al., 2013). For example, B. J. Hale and colleagues (2018) found that when YouTube cancer vloggers explained their diagnosis experience, signaled that they planned to proactively deal with cancer, or provided positive reappraisal of their cancer situation, they were more likely to receive empathic support in comments than vloggers who did not include these narrative elements. In contrast, vlogs that included calls to action—or calls for viewers who are concerned about their health to seek professional attention and care—yielded fewer empathic comments. However, Liu et al. (2013) found that this narrative element promoted vlogger relatability and audience intimacy, which could increase emotional engagement and empathy (Chou et al., 2011). This study provides an opportunity to examine the relationship between these four narrative features and the provision of empathic support within a different platform (i.e., Imgur), and thus the following hypothesis and research question are extended:
H1. The inclusion of (a) the diagnosis experience, (b) problem solving, and (c) positive reappraisal in Imgur posts will increase the likelihood of empathic support in comments compared with posts without these narrative components.
RQ3. Will the inclusion of pleads for audience checkup promote empathic support in comments compared to posts without this narrative component?
In addition, sharing negative experiences (or possible outcomes) may prompt empathy from others. Bareket-Bojmel and Shahar (2011) argue that “negative feelings are important for the processing of the disclosed experience . . . [and] perhaps contribute to the emergence of empathy or closeness among the interacting partners” (p. 752). The salient threat of mortality (to oneself or someone else) is likely to emerge in discussions about cancer, and thus could prompt empathy from other Imgurians. Therefore, it is hypothesized that direct references to mortality (e.g., “the doctors gave me six months to live”) by Imgur posters will elicit empathic reactions from commenters.
H2. The inclusion of mortality in post narratives will increase the likelihood of empathic support in comments compared with posts without mortality disclosures.
Furthermore, previous scholarship has drawn connections between humor, social support, and positive health outcomes (Martin & Lefcourt, 2004; Mora-Ripoll, 2010; Roaldsen et al., 2015). Some scholars argue that humor functions as a distinct coping strategy (e.g., Mora-Ripoll, 2010; Roaldsen et al., 2015), while others contend that humor facilitates support provision (Kuiper & McHale, 2009; Martin & Lefcourt, 2004). Moreover, previous analyses of Imgur content indicate that humor is commonly utilized in this context (Hale, 2017, 2019; Mikal et al., 2014). B. J. Hale (2019) found that non-bona fide linguistic features (e.g., humor, irony, and sarcasm) frequently manifested but rarely coincided with support in comments responding to Imgur posts about depression. However, it is yet unknown whether a relationship exists between posts that utilize humor and the provision of support in comments. Therefore, an additional research question is forwarded to examine this relationship:
RQ4. Will the inclusion of humor in posts affect the likelihood of empathic support provision in comments?
Finally, while previous analyses suggest a prevalence of empathic support online (Blank et al., 2010; Buis & Whitten, 2011; Love et al., 2012), and narrative features seem to best elicit this support type, other forms of support (e.g., reassuring, informational, and tangible) could also relate to narrative components. It may be noted that the previous analysis of YouTube vlogs conducted by B. J. Hale et al. (2018) found that reassuring support was largely unaffected by narrative features, and informational and tangible support types infrequently manifested in YouTube comments, hindering their ability to ascertain the relationship between narrative features and these two support types. However, the nature of supportive communication within Imgur could differ from other platforms (e.g., YouTube), and thus the aforementioned narrative features (i.e., explanations of the diagnosis experience, agentive problem solving, positive reappraisal, and pleads for audience checkup) could predict other forms of support in this context. Therefore, a research question is proposed to examine these possible relationships:
RQ5. Will narrative elements (i.e., the diagnosis experience, pleads for audience checkup, problem solving, positive reappraisal, references to mortality, and humor) predict other types of support (i.e., reassuring, informational, and tangible)?
Method
A content analysis was performed to identify the provision of social support (including reassuring, empathic, informational, and tangible support types) in comments responding to cancer-related Imgur posts. In total, 117 cancer-related Imgur posts and 2511 corresponding comments were analyzed, examining the relationship between post features (i.e., the diagnosis experience, problem solving, positive reappraisal, pleads for audience checkup, and humor) and social support, providing naturalistic insight into online support exchange in Imgur.
Sample
A sampling frame was created by collecting Imgur posts from 9 November 2018 to 12 January 2019. Posts that contained the term “cancer” (i.e., in the title, text, or tags) were identified using the Imgur application programming interface (API) and collected daily following a staggered sampling strategy (N = 241). All user comments submitted to these posts were also collected (N = 8247). Data were collected at 8 a.m. on the first collection day (9 November), and collection was shifted an hour later each day until data were sampled at 8 p.m. (21 November). On the following day (22 November), the schedule began again at 8 a.m., and this process was repeated until 12 January. Screenshots of each Imgur post were captured and stored for analysis. This sampling strategy allowed us to capture content without introducing bias toward posts and comments submitted at a certain time of day or day of the week. From the sampling frame, posts were filtered according to their focus on cancer and the cancer experience (e.g., posts using language like “this movie is cancer” were removed), leaving 122 posts (see Figure 1). Five additional posts that received no user comments were removed, as ascertaining relationships between post and comment categories would not be possible without comments (this decision was made a priori). Thus, the final sample included 117 Imgur posts that directly discussed cancer or the cancer experience and generated at least one user comment. Comments were filtered to only include thread-starting root comments (i.e., comments that directly respond to the post), as focusing on root comments informs our research questions more closely than examinations of response comments that respond to other commenters. A content analysis was performed of the remaining 117 cancer-related Imgur posts (113 unique posters) and 2511 corresponding comments (2322 unique commenters).
A visualization of the filtering process used for posts and comments.
Coding Procedure
A codebook was developed to capture narrative features and control factors in Imgur posts and social support in comments. Two coders familiar with Imgur, including typical posting and commenting practices (especially the use of Imgur-specific humor; see Mikal et al., 2014), were trained using the codebook. Intercoder reliability was assessed using Krippendorff’s alpha (range = .72–1.0). In total, 38 posts (31.2% of the sample) and 480 comments (19%) were examined for coder training and 21 posts (17.2%) and 254 comments (10.1%) were used for calculating Krippendorff’s alpha (overall α = .87). Once reliability was established, the sample was equally distributed between the two coders. Coders accessed each Imgur post via a stored hyperlink or the archived screenshot (if the hyperlink was no longer active) and categorized narrative features (diagnosis, problem solving, positive reappraisal, pleads for audience checkup, mortality, and humor) and control variables (identifiability, cancer type, preexisting support, and cancer subject). Comments were accessed via a spreadsheet and categorized for social support types (reassuring, empathic, informational, and tangible). All content categories were coded across text and visuals (e.g., images or gifs), as multimodal communication is common in Imgur. Problem cases were flagged by coders and collectively discussed.
Measures
Control Categories
Controls, including identifiability, cancer subject, cancer type, and preexisting support were categorized for each Imgur post. Identifiability was constructed by combining three categories related to poster characteristics, including posters that included their image (α = .91), sex (α = .84), and/or age (α = .78). Poster sex was coded as male, female, or unidentifiable, while poster age was grouped into six categories: 0–10 years, 10–20 years, 20–40 years, 40–60 years, 60+ years, or unidentifiable. When the poster remained unidentifiable across these categories, the post was categorized as anonymous. If the poster disclosed their sex or age (or both), this was categorized as minor identifiability. The post was classified as major identifiability when the poster included an image of their self (e.g., a “selfie”). Cancer subject (α = .93) was coded as either self, other person, animal (e.g., pets), or topic (e.g., posts about cancer that did not include a reference to a person or animal). Cancer type (α = 1.0) was categorized according to post content. Finally, preexisting support (α = .92) was categorized as negative support (e.g., negatively affected by family or friends), no support, positive support (e.g., positively affected by family or friends), or very positive support (e.g., emphasizing a particular person’s contribution or a large amount of support), according to information provided by the poster, following the framework used by B. J. Hale et al. (2018).
Narrative Categories
In addition to controls, six narrative categories were coded for each Imgur post, including diagnosis, problem solving, positive reappraisal, pleads for audience checkup, mortality, and humor. Diagnosis (α = .9) was coded using a dichotomous yes/no response for any discussion of cancer diagnosis (e.g., a previous or future diagnosis procedure, diagnosis details, or symptoms that led to diagnosis). Problem solving (α = .79) was also coded using a dichotomous yes/no response and included plans to proactively deal with cancer or concentrated efforts to make things work. Also coded dichotomously, positive reappraisal (α = .77) was categorized when the poster described personal growth or some positive emotional or social change as a result of dealing with cancer. Audience checkup (α = 1.0) was coded when the poster advocated for readers to get a checkup or cancer screening (e.g., a doctor visit for potentially cancerous growths or regular breast examinations). Mortality (α = 1.0) was coded when the poster disclosed a timeline for their possible death, the death of someone else (e.g., family member or friend), or general mortality related to cancer (e.g., statistics about their cancer type). Finally, humor (α = .89) was coded when the poster included jokes, formulaic humor (i.e., Imgur-specific humor), or other humorous language.
Support Categories
Within comments, each support type outlined by Neuling and Winefield (1988) was coded using a dichotomous response. Reassuring support (α = .77) included statements of hope (e.g., “I believe you’ll get through this!”), support (e.g., “you can do this!”), complement (e.g., “you’re looking great!”), and other features designed to uplift the poster. Empathic support (α = .72) was coded when comments contained statements acknowledging the poster’s emotions or feelings (e.g., “I understand where you’re coming from” or “I’ve felt that way too”), or encouragement to continue discussing their experience (e.g., “please keep us updated”). Informational support (α = .83) included verifiable non-personal information, including corrections to misinformation provided by another person. Finally, tangible support (α = .8) was coded when commenters expressed an interest in assisting the poster in a tangible way (e.g., financially). Support types were not mutually exclusive.
Results
Descriptives
Overall, 75.6% of comments contained at least one form of support (N = 1898), suggesting that the typical response to an Imgur post about cancer is supportive (RQ1). Examining specific support types, reassuring support occurred most frequently, followed by empathic, informational, and tangible support (RQ2) (see Table 1). Comments that included informational support received the highest average score of the four support types, followed by empathic, reassuring, and tangible support, and also the highest average number of replies, followed by tangible, reassuring, and empathic support.
Distribution of Support Types in Comments (N = 2511).
Support types were not mutually exclusive. Mean score reports the average score for comments that included the support type. Mean replies indicates the average number of response comments generated by a comment that included the support type.
In addition, each of the narrative features and categories proposed for this study appeared in this dataset (see Table 2). Explanations of the diagnosis experience, agentive problem solving, and mortality were present in approximately 40% of posts, followed in frequency by humor, positive reappraisal, and pleads for audience checkup. A majority of Imgur posters were anonymous, which is unsurprising considering the pseudonymity afforded by the platform (see Table 2). However, posters were identifiable in approximately 44% of cases (N = 52), with nearly 32% of posters (N = 37) providing an image of themselves (i.e., major identifiable). In addition, Imgur posters described the cancer experiences of someone else slightly more frequently than their personal experience, followed by those of an animal (e.g., a pet) or the cancer topic more generally (e.g., cancer statistics, emerging treatments). Approximately 59% of posters (N = 69) detailed their positive support networks, with fewer posters claiming no support network (or neglecting to mention an existing support network), or having experienced negative feedback from their network. Finally, while over a dozen forms of cancer were explicitly described in this dataset, nearly 47% of Imgur posters (N = 55) did not report a specific cancer type.
Distribution of Post Categories (N = 117).
Predicting Social Support
To ascertain the relationship between narrative features and support provision in comments, a series of multilevel binomial logistic regression models were constructed that regressed each support type on narrative and control variables. Models for empathic, reassuring, and informational support were successfully constructed, while tangible support could not be modeled due to scarcity (1.6%, N = 40). Baseline categories were adjusted between support models to assist interpretability, and thus each is reported separately (Tables 3 to 5).
Multi-Level Binomial Logistic Regression Models Predicting Empathic Support.
Baseline categories included anonymous for identifiability, topic for post subject, very positive support for preexisting support, and skin for cancer type. Coefficients for narrative components (diagnosis, problem solving, positive reappraisal, audience checkup, mortality, and humor) represent the change in log-odds when the component is included. Statistical significance is reported according to Wald chi-square. SE = standard error.
p < .1, *p < .05, **p < .01, ***p < .001.
Multi-Level Binomial Logistic Regression Models Predicting Reassuring Support.
Baseline categories included major identifiable for identifiability, topic for post subject, positive support for preexisting support, and other for cancer type. Coefficients for narrative components (diagnosis, problem solving, positive reappraisal, audience checkup, mortality, and humor) represent the change in log-odds when the component is included. Statistical significance is reported according to Wald chi-square. SE = standard error.
p < .1, *p < .05, **p < .01, ***p < .001.
Multi-Level Binomial Logistic Regression Models Predicting Informational Support.
Baseline categories included major identifiable for identifiability, topic for post subject, positive support for preexisting support, and breast for cancer type. Coefficients for narrative components (diagnosis, problem solving, positive reappraisal, audience checkup, mortality, and humor) represent the change in log-odds when the component is included. Statistical significance is reported according to Wald chi-square. SE = standard error.
p < .1, *p < .05, **p < .01, ***p < .001.
Empathic Support
Beginning with the control model (Model 1 in Table 3), post subject emerged as a significant predictor of empathic support (χ2 (df = 3) = 13.17; p < .01), with posters who disclosed the experiences of an animal or another person more likely to receive empathic support in a given comment than those who addressed cancer topically (897% and 540%, respectively). Further pairwise comparisons revealed that posters disclosing their personal experience were also significantly less likely to receive empathic support in a given comment than those addressing the experiences of an animal (p < .02) or another person (p < .02). In addition, cancer type significantly predicted the likelihood of empathic support in comments (χ2(12) = 95.79; p < .001), with cancers of the blood/bone, lung, and esophagus/larynx reporting the highest likelihood of empathic support provision, and skin, pancreatic, and thyroid/endocrine cancers receiving the lowest. Identifiability emerged as nearly significant (χ2(2) = 5.63; p = .06), although pairwise comparisons revealed that minor identifiable posters were 147% more likely to receive empathic support in a given comment than anonymous posters (p < .05). No significant relationship was found between empathic support and preexisting support, (χ2(3) = 2.84; p = ns).
Adding narrative factors to the model yielded a number of additional findings (see Model 2). The first hypothesis predicted that the inclusion of (a) the diagnosis experience, (b) problem solving, and (c) positive reappraisal in Imgur posts would increase the likelihood of empathic support in comments compared with posts without these narrative components. This hypothesis was only partially supported, as explanations of the diagnosis experience increased the likelihood of empathic support in a given comment by 145% (χ2(1) = 9.54; p < .01), while problem solving (χ2(1) = 2.71; p = ns) and positive reappraisal (χ2(1) = 0.77; p = ns) did not significantly predict empathic support. In answering RQ2, pleads for audience checkup did not predict empathic support (χ2(1) = 0.10; p = ns). The second hypothesis predicted that referencing mortality would increase the likelihood of empathic support in comments, and this hypothesis was supported (χ2(1) = 18.08; p < .001). The inclusion of mortality in Imgur posts increased the likelihood of empathic support in a given comment by 250%. Finally, these results indicate a significant negative relationship between humor and empathic support (χ2(1) = 6.27; p < .02), as the inclusion of humor in posts predicted a 56% decrease in the likelihood of empathic support in a given comment (RQ4).
Reassuring Support
Two additional models were constructed to identify the relationship between narrative elements and reassuring support, which emerged more commonly than empathic support in this dataset—in approximately 53% of comments (see Table 1). The control model (Model 3 in Table 4) reported several significant findings, with identifiability (χ2(2) = 9.68; p < .01), preexisting support (χ2(3) = 8.73; p < .04), and post subject (χ2(3) = 8.29; p < .04) predicting reassuring support. Posters categorized as anonymous or minor identifiable were less likely to receive reassuring support in a given comment than those classified as major identifiable (43% and 58%, respectively). Also, posters who reported having no existing support network were 50% less likely to receive reassuring support in comments than those who claimed a positive support network. Posters who disclosed their personal experiences or those of someone else were more likely to receive reassuring support than posters who addressed cancer topically (137% and 153%, respectively). Cancer type did not emerge as significant (χ2(12) = 17.96; p = ns), though pairwise comparisons revealed that some types (i.e., testicular/prostate, esophagus/larynx, and brain) were more likely to receive reassuring support than the baseline category (other cancers).
Adding narrative categories to the model yielded one additional significant finding (see Model 4). Answering RQ5, posters who suggested that the audience should receive a checkup were 48% less likely to receive reassuring support in a given comment than those who did not (χ2(1) = 4.16; p < .05). Diagnosis (χ2(1) = 2.23; p = ns), problem solving (χ2(1) = 1.23; p = ns), positive reappraisal (χ2(1) = 0.91; p = ns), mortality (χ2(1) = 0.00; p = ns), and humor (χ2(1) = 1.26; p = ns) did not significantly predict reassuring support in comments.
Informational Support
Two final models were constructed to evaluate the relationship between narrative elements and informational support. In the control model (Model 5 in Table 5), post subject (χ2(3) = 13.76; p < .01) and cancer type (χ2(12) = 21.5; p < .001) significantly predicted informational support. Compared with posters who discussed cancer topically, those reporting information about themselves were 70% less likely to receive informational support in a given comment, while posters detailing the experiences of another person were 62% less likely, and those describing an animal’s experiences were 67% less likely. Posters describing cancers of the esophagus/larynx, colon, and brain were more likely to receive informational support than those with breast, blood/bone, or thyroid/neuroendocrine cancers. Although preexisting support (χ2(3) = 6.02; p = ns) did not emerge as significant, pairwise comparisons revealed that posters who described negative support from their network were 151% more likely to receive informational support than those reporting a positive support network (p < .04). Moreover, posters who described no support network were 65% more likely to receive informational support than those with a positive support network, although this was nearly significant (p = .09). Identifiability (χ2(2) = 4.57; p = ns) was not significant.
When narrative categories were included (Model 6), positive reappraisal (χ2(1) = 8.2; p < .01) emerged as a significant predictor of informational support (RQ5). Posters who provided a positive reappraisal of their situation were 168% more likely to receive informational support in a given comment than those who did not include this feature. Explanations of the diagnosis experience (χ2(1) = 0.49; p = ns), problem solving (χ2(1) = 1.45; p = ns), audience checkup (χ2(1) = 0.1; p = ns) mortality (χ2(1) = 0.00; p = ns), and humor (χ2(1) = 0.48; p = ns) were not significant.
Discussion
Through a content analysis of cancer-related Imgur discourse, this project examined the relationship between narrative features in 117 cancer-related Imgur posts and support provision in 2511 corresponding comments. Building upon previous social support work (Chou et al., 2011; Dunkel-Schetter et al., 1987; Hale et al., 2018; Kuiper & McHale, 2009; Liu et al., 2013; Martin & Lefcourt, 2004), six narrative features were examined, including explanations of the diagnosis experience, agentive problem solving, positive reappraisal, pleads for the audience to get a checkup (e.g., cancer screening), references to mortality, and humor. In addition, four types of social support were identified, as outlined by the MDSS (Neuling & Winefield, 1988)—reassuring, empathic, informational, and tangible support. This project responds to recent calls for research evaluating antecedents to support provision and support emergence across social media platforms (Meng et al., 2017; Rains et al., 2015), finding that the inclusion of narrative elements in cancer-related Imgur posts significantly predicted the generation of support in responding comments. Because the Imgur platform differs culturally and architecturally from other well-researched sites like YouTube (Frohlich & Zmyslinski-Seelig, 2012; Hale et al., 2018; Liu et al., 2013) and Facebook, which has received the bulk of previous scholarly attention (Buehler et al., 2019; Gage-Bouchard et al., 2017; High et al., 2014; Vitak & Ellison, 2013), the results of this study are well-positioned for comparison with previous findings.
The results of this study suggest that the typical response to an Imgur post about cancer is supportive, as approximately 75.6% of comments in this dataset contained at least one form of support. Reassuring support manifested most commonly (in approximately 53% of comments), followed by empathic, informational, and tangible support types. This pattern is consistent with that found by B. J. Hale et al. (2018), although reassuring and empathic support occurred slightly less frequently and informational support emerged more frequently in Imgur than in YouTube. These findings also agree with previous support research indicating increased provision of empathic compared to informational support in response to cancer patients in other online communities (Blank et al., 2010; Buis & Whitten, 2011; Love et al., 2012). While reassuring support emerged more prominently than empathic and informational support in both Imgur and YouTube comments (Hale et al., 2018), this support type has been largely excluded from previous online cancer-related social support research (see Rains et al., 2015), and thus the relative prevalence of this support type is difficult to assess across studies. In addition, the findings of this study differ notably from a previous analysis of depression-related Imgur posts (Hale, 2019), where reassuring support emerged at a similar frequency to empathic and informational support (and lower levels than that found here), followed by tangible support. Considered together, these results agree with the propositions of optimal matching theory (Cutrona & Russell, 1990; Rains et al., 2015), as stressor characteristics (i.e., cancer versus depression) seem to affect the provision of different forms of support more than the platforms through which support is sought and provided. Moreover, considered alongside previous scholarship these findings indicate that empathic and informational support is generally available across digital platforms (Rains et al., 2015; Zhang et al., 2017). As an additional note regarding the 24.4% of comments that did not contain social support, these often included dark humor (e.g., “enjoy death!”), mean-spirited or hateful content (e.g., “Buck up and get over it”), questions (e.g., “CT stands for cancer tumors?”), and irrelevant content (e.g., “Can we deploy it against Facebook?”), and were typically poorly received by the community.
As hypothesized, Imgur posters who included an explanation of their diagnosis experience were more likely to receive empathic support from commenters than those who did not. Unexpectedly, however, agentive problem solving and positive reappraisal did not predict empathic support in Imgur comments. Furthermore, pleads for the audience to receive a professional checkup did not relate to empathic support. These results differ from previous work (Chou et al., 2011; Dunkel-Schetter et al., 1987; Hale et al., 2018; Liu et al., 2013) and could be the result of differences in audience attention and expectations of message detail, where Imgur posts are generally brief and humorous (promoting browsing), while YouTube vlogs are lengthier and audiences are expected to remain engaged for a comparatively longer amount of time. These narrative features may have encouraged supportive interaction through the YouTube platform where detail is comparatively more prioritized than Imgur. Another possibility is that these narrative features are better expressed to an audience familiar with the cancer patient’s situation (e.g., followers, subscribers, or an offline network), which is a core characteristic of platforms like Facebook and YouTube, but not central to Imgur. Agentive problem solving and positive reappraisal indicate a positive or healthy transformation in the poster’s mindset (e.g., “I have decided to take charge and try something new”; “I have now developed a new outlook on life”) and thus could have more impact for familiar and intimate audiences, which are more likely to provide empathic support (Chou et al., 2011).
Also as hypothesized, references to mortality significantly increased the likelihood of empathic support in Imgur comments. Mortality is a regularly articulated concern for posters (this feature emerged in nearly 40% of posts), and sharing fears associated with impending mortality may facilitate empathy and emotional closeness with the audience (Bareket-Bojmel & Shahar, 2011), increasing the likelihood that commenters will provide empathic support. In addition, the inclusion of humor significantly decreased the likelihood of empathic support in comments. One possibility is that the use of humor by posters prompted humorous (instead of supportive) responses, reframing the cancer discussion within normative (i.e., humorous) Imgur discourse (see Mikal et al., 2014). Put differently, while other narrative features prompted empathic responses, humor may have signaled different psychosocial needs (e.g., conversational normalcy) to commenters. However, the causal relationship between post-level humor and comment-level support cannot be ascertained from these data.
Although less sensitive to narrative features than empathic support, two significant findings emerged for reassuring and informational support. First, Imgur posters who suggested that the audience should receive a professional checkup (e.g., cancer screening) significantly decreased their likelihood of receiving reassuring support. Perhaps this suggestion was experienced as an unwelcome imposition (maybe similar to a public health message), especially from an unknown, and potentially anonymous source, hindering commenters’ willingness to provide reassuring support. Alternatively, the inclusion of this narrative feature may have shifted the discourse away from the poster and onto commenters, reducing the number of comments providing reassuring support. Second, posters who positively reappraised their situation increased their likelihood of receiving informational support. Speculatively, one possibility is that individuals felt more comfortable providing informational support to the poster when they provided a positive outlook, potentially indicating a receptiveness to advice (or perhaps the absence of positive reappraisal indicates a lack of receptiveness). However, the causal relationship between these narrative features and support should be explored in future work.
The relationship between two control categories—identifiability and post subject—and support provision are also worth expounding. Posters who included “major identifiable” features in posts experienced the highest likelihood of empathic and reassuring support, but the lowest likelihood of informational support. Conversely, anonymous posters were the least likely to receive empathic support, while posters with fewer identifying characteristics received the lowest level of reassuring and highest level of informational support. In other words, differences in support are contingent on identity, a peculiar predictor of supportive engagement, particularly in digital platforms that host cultures of anonymity like Imgur and Reddit (Brown et al., 2018). Voluntary disclosure of identifying information seems to have important implications for support provision. In particular, posters who relinquished their anonymity by sharing an identifiable image seemed to benefit from increased empathic and reassuring support. Next, posters who described the cancer experiences of an animal were significantly more likely to receive empathic support than those describing their own personal experience, those of another person, or cancer topically. Disclosing personal cancer experiences or details about another person yielded the highest levels of reassuring support, while discussing cancer as a topic generated the most informational support. Thus, the relationship between the poster and cancer (e.g., patient, family member, pet owner) seems to have further ramifications for support provision. Considering the importance of empathic support for positive health outcomes (Iso-Ahola & Park, 1996; Uchino et al., 1996), it is noteworthy that posters describing the experiences of an animal were most likely to receive this support type. Perhaps the frequency of cancer in domestic animals (American Veterinary Medical Association, 2019) and the increased likelihood that commenters shared the poster’s experience facilitated empathic responses.
Overall, the results of this study indicate a relationship between narrative construction and support provision, finding that the inclusion of narrative features in cancer-related Imgur posts influenced the provision of social support in comments. Moreover, consistent with previous scholarship, narrative features primarily predicted empathic support (Chou et al., 2011; Dunkel-Schetter et al., 1987; Hale et al., 2018; Liu et al., 2013). Accordingly, this work contributes to social support research by (1) identifying the supportiveness of Imgur comments generated in response to messages about cancer, providing insight into a popular social media platform that has received little previous scholarly attention, (2) examining the relationship between narrative features in cancer-related Imgur posts and support provision in comments, and (3) allowing comparisons of support generation between Imgur and previously studied sites.
Findings of this study could have implications for a multitude of stakeholders interested in social support provision, including healthcare professionals, researchers interested in the use of social media platforms for support, and organizations interested in designing supportive online platforms for individuals coping with cancer. First, these results indicate that supportive feedback about cancer is similar across online social networks (e.g., Facebook, YouTube, and Imgur). Recognizing differences in cultural expectations and platform affordances, alongside cancer patients’ existing social media presence, may allow for more specific recommendations about where patients could find support online. For example, practitioners and caregivers might direct individuals coping with cancer to Imgur who would benefit from the site’s distinct functions, including anonymity and humor-based communication. Patients and caregivers that have social anxiety or prefer to keep their health condition private might find that Imgur is a comfortable place to connect with others and acquire support. Second, organizations that advocate, fundraise, and create communities for cancer patients and survivors (e.g., Susan B. Komen Breast Cancer Foundation and Livestrong) often have robust communities on larger social media sites like Facebook and Twitter. By acknowledging and building communities in anonymous networks like Imgur, these organizations could reach broader audiences and better cater to patient personalities and preferences. Finally, similar social support findings between architecturally dissimilar sites like Imgur (a centralized content-oriented pseudonymous platform), YouTube (a decentralized “channel”-based platform with varying levels of anonymity), and Facebook (a decentralized user-oriented identifiable platform) suggest generalizable patterns of supportive communication, which is likely of interest to social support researchers. In particular, the provision of empathic and informational support to cancer patients and the sensitivity of empathic support to voluntary self-disclosure and narrative elements seems consistent across contemporary digital platforms.
Limitations and Future Directions
A few limitations of this study and opportunities for future research should be noted. Although the results of this study indicate significant relationships between narrative features in Imgur posts and social support in comments, causality cannot be ascertained from content analysis data. Nevertheless, the results of this study set the foundation for further exploration of support provision in niche social networks, including the use of narrative features for support elicitation, as this study reinforces previous scholarship suggesting that narrative construction can influence support. Additional examinations of cancer narratives will provide important points of comparison, especially within other social media platforms (e.g., Reddit, Tumblr, Instagram). Specific attention to differences in visual and textual communication might elucidate important distinctions between support communities and social media cultures, especially comparing platforms that rely heavily on visual communication (e.g., Imgur and Instagram) to those more narratively focused. As shown in the findings of this research, further work within diverse platforms may elucidate differences in narrative construction and supportive communication across digital contexts, and thus should prove valuable for our understanding of online support generation.
Conclusion
This study suggests a relationship between narrative construction and social support, as the inclusion of narrative features in cancer-related Imgur posts influenced the provision of support in responding comments. Developing a robust support network (including online connections) and increasing perceived social capital helps individuals battling diseases and health conditions like cancer facilitate adaptive coping mechanisms. Accordingly, understanding what factors facilitate the provision of support in online conversations could help researchers, healthcare professionals, and platform designers develop spaces and strategies for eliciting support and guiding productive conversation about cancer. Furthermore, exploring supportive communication in small social networks like Imgur, dwarfed by well-studied platforms like Facebook and YouTube, provides opportunities to examine both differences and similarities in discussions about cancer across social media communities. Despite the distinct differences in peer-to-peer (e.g., Facebook) and content-based (e.g., Imgur) networks, this study confirms that the communal response to people posting about cancer tends to be supportive. Future work should take account of the growing diversity of online communities, including the connections and support emerging across contemporary social media platforms.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
