Supporting self-efficacy through interactive discussion in online communities of weight loss

Abstract

By conducting a content analysis of online communities connected by the Weight Watchers’ online message boards, this study examined the relationship between conversational interactivity and consumer-generated content about consumer health information, self-efficacious content, and experiences with dieting and physical activities. The results showed that discussion about successful experiences with weight loss tended to be more interactive. Discussion about consumer health information tended to be non-interactive. The findings suggest that online communities generate social support through interactive discussion about successful experiences, and the interactive discussion, in return, sustains active participation in the community.

Keywords

conversational interactivity online health communities self-efficacy social media weight loss

Introduction

Online weight loss communities have grown increasingly popular as overweight and obesity have become more prevalent (Chang et al., 2013). The Centers for Disease Control and Prevention (2014) report that more than one-third of adults are overweight (33.3%) and one-third are obese (35.7%). As a result, there has been an increasing demand for health information, and weight loss products and services. For example, 108 million Americans dieted in 2012, and 70 percent of these dieters used commercial diet plans such as Weight Watchers or Jenny Craig (Marketdata, 2013). Many Americans (60%) also engaged in tracking health behaviors such as weight, dieting, or exercise routines (Pew Internet & American Life Report, 2014). As consumers take more of an interest in their health, they often go in search of information on the Internet; in fact, 72 percent of Americans looked online for health information last year (Pew Internet & American Life Report, 2014). Eighty-four percent of Internet users participated in some kind of online communities, and roughly one-third of these communities were related to health (Jones and Fox, 2009). The social support and feelings of interconnectedness derived from engaging with like-others on social media is a promising resource, especially with regard to reducing overweight and obesity (Chang et al., 2013).

Research on social media and its impact on weight loss is still a burgeoning area of study (Hwang et al., 2010; Napolitano et al., 2013). Studies generally found that people use social media, including weight loss social media sites, Facebook, online discussion boards, to exchange social and informational support, and share personal stories (Ballatine and Stephenson, 2011; De la Peña and Quintanilla, 2015; Liang and Scammon, 2011; Wang et al., 2014). Others indicated that social media play a role in retaining and engaging users’ participation (Gil de Zúñiga et al., 2012; Kiesler et al., 1984; Thackeray et al., 2003). However, the wide variety of social media involves a range of consumers’ engagement activities, for example, commenting, sharing, and liking (Kaplan and Haenlein, 2010; Muntinga et al., 2011). Arguably, a better approach is to examine each type of social media in order to understand the relationship between consumers’ engagement activities and weight loss.

Among various consumers’ engagement activities, textual discussion places conversational interactivity at the center of users’ engagement and participation. Successful online discussion should facilitate interactive conversation among users (Rafaeli and Sudweeks, 1997). Conversational interactivity, thus, is defined in the literature as a contingent process of conversation that involves exchanges of messages semantically relevant or irrelevant to each other (Harvey-Berino et al., 2004; Rafaeli and Sudweeks, 1997). In practice, efforts have been made to explore message strategies to increase conversational interactivity and sustain participation in online communities of physical activities (Resnick et al., 2010). These efforts point to an urgent need to understand the relationship between content features of the discussion and users’ participation.

Several studies (Liang and Scammon, 2011; Wang et al., 2014) suggest that health communication professionals learn from successful peer-to-peer online communities, for example, Weight Watchers’ online communities. These consumer-oriented online communities often have a larger membership and more active participation than online health interventions. Thanks to rich textual discussion of these online communities, researchers can not only examine the content being exchanged (e.g. see Liang and Scammon, 2011; Wang et al., 2014) but also make connections among aspects of communication. Following this line of research, this study analyzes the connection between conversational interactivity and content features that relate to weight loss behaviors, inclusive of consumer health information, self-efficacious content, dieting, and physical activities. Self-efficacious content refers to content addressing the sources of self-efficacy—a person’s belief in his or her ability to successfully perform a specific behavior (Bandura, 1991, 1993).

Literature review

Social media is a broad term. It refers to Internet-based applications that allow the creation, exchange, and sharing of user-generated content (UGC) (Kaplan and Haenlein, 2010). Under this general definition, there are various types of social media (Kaplan and Haenlein, 2010). Some types of social media provide rich media for users to create acoustic, visual, and other sensory contacts with others (Kaplan and Haenlein, 2010). For example, Secondlife.com allows users to create an animated avatar to interact and socialize with other avatars in a virtual world. Other types of social media are used to create and control a desirable self-image (Kaplan and Haenlein, 2010). For example, Facebook is often used to document one’s life to present a desirable self-image. Thus, each social media site is different depending on the extent to which it can be used for social presence, that is, creating acoustic, visual, and other sensory contacts, or for self-presentation, that is, creating and managing a desirable self-image (Kaplan and Haenlein, 2010). Thus, the diverse types of social media require an extensive examination of various social media sites in relation to health issues.

Earlier studies on social media in health communication examined online discussion boards, for example, breast cancer discussion boards (Rodgers and Chen, 2005). With the rise in social networking sites, the focus of research turns to health social networking sites, for example Dailystreght.org, and found that health social networking sites contained rich content on informational and emotional support and personal experiences (Liang and Scammon, 2011). Because of its popularity, Facebook is used for health purposes. Applying netnography, De la Peña and Quintanilla (2015) found that people use Facebook to obtain encouragement and health information and share success stories, which in turn motivate others to achieve health goals. Meanwhile, Facebook pages of commercial weight loss programs, for example, Weight Watchers, are subjected to analysis due to the program’s alleged effects and large number of consumers. By surveying users’ perception and their browsing and commenting behavior, a study identified different types of users on Weight Watchers’ Facebook page and differentiated the levels of informational and emotional supports that they receive (Ballatine and Stephenson, 2011).

Built upon these previous studies, this research examines UGC in an online community connected by the online discussion boards of Weight Watchers. Online communities connected by online discussion boards are different from Facebook and other social networking sites since online discussion boards are primarily structured around topical threads not individual profiles. Thus, online communities connected on discussion boards are largely built upon textual discussion. Accordingly, users’ activities focus less on managing self-presentation but more on exchanging content and opinions on issues. As a result, conversational interactivity of the textual discussion is key to users’ participation and engagement (Wang et al., 2015). Simply, the more interactive the conversation, the more active users’ participation will be. Meanwhile, the more frequently members exchange health-related content, the more widespread such content is and the greater impact it may have. Thus, to advance the understanding of social media in weight loss, this study analyzes conversational interactivity and content features of UGC occurring in the online community of Weight Watchers’ discussion boards.

Conversational interactivity

Conversational interactivity originates in interpersonal communication (Rafaeli and Sudweeks, 1997). It views interactivity as a contingent process of conversation (Harvey-Berino et al., 2004). A theory of conversational interactivity is Rafaeli’s (1988) responsiveness model of interactivity. According to this model, there are three levels of human interaction: non-interactive, reactive, and interactive. Non-interactive conversations are bilateral flow of messages that do not refer to or cohere with each other. Reactive conversations require later messages refer to or cohere with earlier messages in a message flow. Interactive conversations require that later messages not only semantically cohere with earlier messages but also indicate how earlier messages are interconnected. Thus, Rafaeli’s responsiveness model of interactivity is largely based on semantic relatedness of a conversation (Bucy and Tao, 2007); messages share meaning and semantic associations in a reactive or an interactive conversation.

Consumer health information

The first type of UGC that this study examines in relation to conversational interactivity is consumer health information. Consumer health information refers to facts that enable individuals to understand their health and make health-related decisions (Patrick and Koss, 1995). For example, information on calories in diets and drinks, exercise videos, and online and offline weight loss services is exchanged in online weight loss communities (Wang et al., 2014). Consumer health information can reduce uncertainty regarding health status and construct a social and cognitive sense of health (Tardy and Hale, 1998). Consumer health information on benefits, barriers, severity, and susceptibility of health conditions can facilitate behavior change (Rimer and Glanz, 2014). People obtain consumer health information from medical professionals such as doctors, mass media such as magazines and newspapers (Cotton and Gupta, 2004), and interpersonal networks, such as friends and family (Tate et al., 2003). The Internet is also an important source of consumer health information (Ginossar and Nelson, 2010). Pew Research and Internet Project showed that 59 percent of US adults reported searching for consumer health information online in the past year (Fox, 2013).

Because of the importance of consumer health information, studies explored ways of using web-based interactivity to create engaging, user-friendly, adaptable, and accessible content that communicates and promotes health (Neuhauser and Kreps, 2010). By analyzing 100 health websites, including .com, .org, .gov, and .edu (e.g. http://www.nimh.nih.gov), McMillan et al. (2008) found that human-to-computer interactive features are more frequently located on health websites than human-to-human and human-to-content interactive features. Another content analysis revealed that interactive elements are not frequently found on the 157 breast cancer websites sampled (Whitten et al., 2008). These content analyses urged the application of interactive communication strategies on health-related websites, for example, live consumer support, options for emailing content to a friend, and customized content (McMillan et al., 2008).

However, conversational interactivity in online communities is mostly based on human-to-human communication, and thus is quite different from interactivity on websites. Conversational interactivity in online communities depends on participants’ responsiveness to discussion topics. While studies generally suggested that consumer health information is a main topic among consumers on social media (Ballatine and Stephenson, 2011; De la Peña and Quintanilla, 2015; Liang and Scammon, 2011; Wang et al., 2014), there is no empirical evidence regarding whether consumer health information is a topic that members interactively respond to in an online conversation.

Self-efficacious content

The second type of UGC this study examines in relation to conversational interactivity is self-efficacious messages. Self-efficacious content refers to posts containing successful personal experiences with weight loss, positive feelings about those experiences, and encouragement to stay on track and reach weight loss goals. According to Bandura (1991), a person’s self-efficacy can be influenced by observing others’ successes with the behavior or by encouragement. In an online community, a person can learn about another member who succeeded at a behavior from the personal stories shared on social media (Willis, 2016). As a result, social comparisons of oneself to like-others with regard to particular behavior can strengthen one’s own efficacy of performing similar behaviors (Bandura, 1991). In other words, observing other people’s behaviors allows for a safer and more efficient way of learning behaviors rather than trying and failing. Bandura also addressed the importance of social persuasion in enhancing self-efficacy. Social persuasion refers to the encouragement one receives from support systems. Finally, affective factors also influence behaviors; for example, happiness or anxiety. Based on Bandura’s (1991, 1993) theory, this study focuses on three types of self-efficacious content: (1) successful personal experience with weight loss, (2) encouragement, and (3) affect associated with weight loss. The existing empirical evidence supports the existence of these three types of self-efficacious content in online health communities (Wang et al., 2014; White and Dorman, 2001). However, the relationship between self-efficacious content and conversational interactivity has not been adequately addressed. In theory, self-efficacious content can potentially be interactive since people can easily relate to like-others’ experiences and/or the feelings of weight loss. Through research, this study provides empirical evidence for the relationship between self-efficacious content and conversational interactivity.

Dieting and physical activities

Finally, consumer-generated content about dieting and physical activities may also be associated with higher interactivity in an online weight loss community. In 2002, Burgoon et al. argued that computer-mediated communication contains fewer social and contextual cues than interpersonal communication and thus shifts attention to the task(s) under discussion rather than other community members. Today, this statement is still applicable to online discussion boards, where users’ interaction focuses less on self-presentation. In online communities connected by discussion boards, community members “tune in” to relevant topics of weight loss, such as dieting or exercise, and give more cognitive resources to understanding these tasks under discussion (Dutta-Bergman, 2004). Thus, online discussion relevant to dieting and physical activities may also be content that people would like to interact with and thus associated with conversational interactivity.

Research questions

The above review of literature identifies three types of UGC that relate to weight loss: consumer health information, self-efficacious content, and content about dieting and physical activity. First, to confirm the occurrence of these types of content, this study examines the following research question:

RQ1. How frequently do content about physical activities, dieting, consumer health information, successful personal stories, encouragement, and affect appear in discussion of an online weight loss community?

Second, to investigate the association of these types of content with conversational interactivity, this study examines the following research question:

RQ2. Which content features are associated with higher conversational interactivity in an online weight loss community, physical activities, dieting, consumer health information, successful personal stories, encouragement, or affect?

Method

Sampling

The online communities connected by Weight Watchers’ online message boards were chosen for this study. Weight Watchers is one of the largest American-based commercial weight loss programs that attract a large membership with high frequencies of participation. Content analysis was used as the method to examine textual discussions occurring within the online communities of Weight Watchers’ message boards. The sampling occurred between 1 December 2010 and 31 January 2011—a time period when people might worry about their weight due to the major holidays, had tried to diet, and sometimes made New Year’s resolutions to lose weight. Thus, it is likely to gather a sample of discussion with a higher relevance to the topic of weight loss.

Given the nature of the data, this study was exempt from institutional review board (IRB) review where this study was conducted for the following reasons: first, the collected conversations took place entirely in the public domain. The message boards of Weight Watchers did not require registration to view the discussion. According to Brem (2002), research is exempt from the protection of human subjects when people know they are publishing to a public area with unrestricted viewing. Second, the data collection of this research does not involve intervention or interaction with living individuals. Third, no information that can identify any individual participants was reported in this article, such as user names, participants’ signatures, and email addresses.

To keep the context of the bilateral flow of discussion, the sampling procedure took “snap-shots” by selecting consecutive posts from randomly sampled threads rather than selecting random posts. There were three steps to the sampling: (1) eight sub-forums were randomly selected from 58 sub-forums, (2) a random sampling of threads within each selected sub-forums was conducted, and (3) consecutive posts, starting with the most recent one, were selected within the sampled timeframe from each thread. Six was set as the maximum number of consecutive posts from each thread based on the researchers’ judgment and the feasibility of the coding. Thus, if the thread had more than six posts in the sampled timeframe, only the most recent six posts were selected. If the thread had fewer than six posts in the sampled timeframe, all consecutive posts were selected. Using this sampling procedure, this study analyzed 400 posts from 78 threads. Each sampled thread had an average of 5.12 posts.

Coding categories and intercoder reliability

The unit of coding was each post in order to capture targeted content features. Conversational interactivity is, however, a feature of conversational threads. To capture levels of conversational interactivity, two mechanisms were put into work. First, the snap-shots sampling procedure made sure that consecutive posts in a thread were captured. Second, the definition and operationalization of “interactivity” applied in the coding required comparison of a post to other posts in the same thread to determine its level of interactivity. Thus, the coding was able to reflect levels of interactivity in a conversational thread.

Two coders underwent several rounds of training, which included explaining the codebook, demonstrating the coding, and practicing the coding. When the coders exhibited sufficient mastery of the coding, they independently coded 10 percent of the sampled posts, which were used to calculate intercoder reliability. Cohen’s Kappa was calculated for all coding variables except conversational interactivity. An average Cohen’s Kappa coefficient of 0.84 was achieved, and the minimum requirement of 0.75 was met for every variable coded (Trochim, 2015). As a higher level of conversational interactivity had an additional feature than the lower level, it was treated as a Guttman scale, a type of ordinal measure (Trochim, 2015). Therefore, Krippendorff’s Alpha was used to indicate the reliability of an ordinal variable. The Krippendorff’s Alpha showed that the intercoder reliability of conversational interactivity was 0.91 (Hayes and Krippendorff, 2007). After all variables met the minimum requirement of intercoder reliability, each coder coded 50 percent of the sampled posts.

To identify threads and posts, the following information was coded:

ID of the thread. Each thread excerpt had a unique ID number;

Order in the thread. The order of the post as it appeared in the thread excerpt;

Screen name of the author. Coders wrote down the screen name of the author.

To measure variables, the following items were coded into binary values: 0 = absent, 1 = present.

Consumer health information

Whether factual information regarding health risks and access to appropriate health services is mentioned. For example, a participant pointed out that a type of cocktail had a lot of calories; several other participants discussed dieting with irregular bowel syndrome. Intercoder reliability was 0.95.

Physical activities

Whether the post mentioned physical activities from a weight loss perspective. For example, the post reported the author’s daily exercise routine saying: “I did an old VHS tape today … I really sweat during this one and today was no exception.” Intercoder reliability was 0.78.

Dieting

Whether the post mentioned dieting from a weight loss perspective. For example, the post reported the author’s daily food intake saying, “Had only one portion of my red light food—spaghetti and white clam sauce. I usually go back for fifths!” Intercoder reliability was 0.9.

Successful personal experience with weight loss behavior

Posts that described success of personal experience with certain weight loss behaviors. For example, the post reported that the author successfully resisted food temptation or finished exercise. The two examples in physical activities and dieting were also examples of successful experiences with weight loss behavior. Intercoder reliability was 0.85.

Encouragement

Encouragement of health behaviors, with the intention to change or reinforce similar-other’s health behaviors, including posts that simulated weight loss behavior by approval or appraisal. For example, “Hi, Mary. Good job walking the stairs this morning! Let’s try for an extra flight of stairs tomorrow.” Intercoder reliability was 0.77.

In addition, Affect had four categories and Conversational Interactivity had three categories.

Affect was defined as feelings or emotions related to weight loss. It was operationalized as whether the post-contained words and language that express feelings, such as anger, embarrassment, and joy. Intercoder reliability was 0.86. The values of Affect included the following:

The affect was neutral when no feelings or emotions were revealed.

The affect was negative when words such as sad, angry, and frustrated, and similar language were used to express unpleasant feelings.

The affect was positive when words such as excited and happy, and similar language such as smiley faces were used to express pleasant feelings.

When a post revealed both positive and negative affects, it was coded as mixed affect.

Each post was assigned one of the four Affect values.

Conversational interactivity was defined as responsiveness of message exchanges. Based on Rafaeli’s (1998) framework, conversational interactivity had three values: non-interactive, reactive, and interactive message exchanges (intercoder reliability = 0.91.). Each post was assigned each of these three values:

Non-interactive message exchanges were operationalized as posts that were not coherent with the topic in any of the posts in a thread excerpt.

Reactive message exchange was operationalized as posts that referred to previous posts or were coherent with the topics of a previous post in a thread excerpt involving at least two different members.

Interactive interaction was operationalized as posts that possessed all the qualities of reactive posts, and also addressed the exchange of conversation, or the relationship between previous posts or questions and answers in previous posts. In other words, an interactive post was related to multiple previous posts.

For example, the third post in the following discussion is considered interactive since it is coherent with the topics in posts 1 and 2 and refers to the conversation on Weight Watchers’ point-friendly gluten-free bread and Vietnamese rice wrappers in the previous two posts:

Post 1 by Member A: I have been GF [gluten-free] for 6 years now. I sure wish I could find some good low point GF bread. Sometimes a sandwich is what I am craving. Just thought I would stop in and say hello!

Post 2 by Member B: I’ve never had any luck on the gluten-free bread front; however, I do like Vietnamese rice wrappers. They are like a big noodle. I build my sandwich inside. I double them up when I don’t eat my sandwich right away. They are pretty delicate and tear easily. I’m adding the link so you can see what they look like: [a URL link]

Post 3 by Member A: Thanks! I will look for those. I’m guessing they are fairly point-friendly?

Results

RQ1 asked the occurrence of content features on the message boards. The results for RQ1 are reported in Table 1. The results confirmed the existence of consumer health information, self-efficacious content, dieting, and physical activities in the online discussion.

Table 1.

Frequencies of health-related content.

Coded items	Categories	Frequencies (%)
Consumer health information	Presence	53 (13.2)
Physical activities	Presence	164 (40.8)
Dieting	Presence	105 (26.1)
Successful personal stories	Presence	114 (28.4)
Encouragement	Presence	99 (24.6)
Affect	Neutral	245 (61)
	Positive affect	143 (35.6)
	Negative affect	11 (2.7)
	Mixed affect	3 (0.7)
	Total	402

A closer examination of consumer health information showed that members shared websites with suggestions for abdomen crunches and other exercises, recipes, and snack ideas for those with dietary restrictions, for example. Members also shared recommendations for mobile health applications and gluten-free or dairy-free products. Of the posts that included consumer health information, many were related to other health issues such as diabetes and hypoglycemia perhaps caused by overweight and obesity.

RQ2 investigated the relationship between content features and interactivity. Descriptive statistics showed that non-interactive (N = 127), reactive (N = 142), and interactive posts (N = 133) were almost evenly distributed in the sample. To answer RQ2, logistic regressions were conducted. The dependent variable was the level of interactivity and the independent variables were physical activities, dieting, health information, successful personal stories, encouragement, and positive emotion, all of which were dummy-coded. Negative and mixed emotions were excluded because of extremely low frequencies.

To build models for the relationship between the dependent variable and independent variables, the statistical analysis started with a multinomial regression model. In this model, the dependent variable was the level of interactivity, which had three values, non-interactive, reactive, and interactive. The multinomial regression model was significantly different from the intercept-only model (χ² = 58.61, p < 0.00, pseudo R² = 0.15), but failed the goodness-of-fit test (χ² = 110.96, p < 0.05). Thus, the multinomial regression model was not a good fit for the data.

Thus, this study applied binary logistical regression to build three separate regression models rather than one multinomial regression model. Accordingly, three binary logistic regressions were conducted for simple comparisons between levels of conversational interactivity.

In the first binary logistics regression, the dependent variable had two levels: non-interactive (N = 130) versus reactive (N = 134). All interactive posts were excluded. The results showed that the regression model was significantly different from the intercept-only model (χ² = 36.12, p < 0.00). However, the goodness-of-fit test showed that the model prediction was significantly different from the observed values (χ² = 19.10, p < 0.00). Thus, the reported model was not a good fit for the data. The observed difference between non-interactive and reactive posts in terms of consumer health information, physical activities, and encourage was not supported by the empirical data.

In the second binary logistic regression, the dependent variable had two values: interactive (N = 136) versus non-interactive (N = 130). Reactive posts were excluded. The results showed that the regression model was significantly different from the intercept-only model (χ² = 39.29, p < 0.00). The pseudo R² was 0.18. The goodness-of-fit test showed that the model was not significantly different from observed values (χ² = 11.91, p = 0.064). Thus, the reported model was a good fit for the data. The results showed that a unit’s increase (from 0 = absent to 1 = present) in consumer health information (β = −1.06, Exp(B) = 0.35, p < 0.01), physical activity (β = −0.66, Exp(B) = 0.52, p < 0.05), and encouragement (β = −1.04, Exp(B) = 0.36, p < 0.01) decreased the likelihood of being an interactive post. A unit’s increase (from 0 = absent to 1 = present) in successful experience (β = 1.36, Exp(B) = 3.90, p < 0.00) increased the likelihood of being an interactive post. These results showed that a post was more likely to be interactive rather than non-interactive when successful experience was present. A post was more likely to be non-interactive rather than interactive when consumer health information, physical activities, or encouragement was present in the post.

In the third binary logistic regression, the dependent variable had two values: interactive posts (N = 136) versus reactive posts (N = 134). Non-interactive posts were excluded. The results show that the regression model was not significantly different from the intercept-only model (χ² = 8.69, p = 0.19). None of the independent variables was significant. The results indicated that there was no significant difference between interactive posts and reactive posts in terms of content features.

Discussion

The role of the online community

The binary logistic regression revealed that the presence of successful experience with weight loss increased the chance of being an interactive post rather than a non-interactive post. This finding suggests that participants actively commented on others’ successful experiences with weight loss. Possibly, participants could easily relate to and/or congratulate others’ weight loss experiences. Thus, theoretically, it is suggested that online communities connected by discussion boards facilitate interactive discussion about self-efficacious content, which is an important source of social support. Practically, this finding suggests sharing successful experiences can be used as a content strategy to encourage participation and generate social support. Actually, some online weight loss interventions have begun intentionally applying content strategies (Gold et al., 2007). For example, Stepping Up to Health, an Internet-mediated walking program asked their moderation staff to share their own experiences with physical activities in order to sustain active participation in the online community (Gold et al., 2007). All in all, discussion about successful personal experiences can be strategically used to increase conversational interactivity, which, in return, sustains social support and participation in online communities of weight loss.

The findings also revealed that consumer health information was negatively associated with conversational interactivity. The binary logistic regression showed that the presence of consumer health information in a post increased its chance to be a non-interactive post rather than an interactive post. Theoretically, this finding suggests that members of this particular all-peer online community might not actively participate in discussion on consumer health information. A possible explanation is that having an interactive discussion on consumer health information is limited by community members’ varying knowledge levels. As suggested by studies on health literacy, it has been challenging for consumers to digest consumer health information online (Benigeri and Pluye, 2003; Birru et al., 2004; Cline and Haynes, 2001).

Practically, this finding presents both challenges and opportunities for health educators. It is an opportunity for health educators to provide consumer health information through the networked online community. An active online community with a large membership resembles a media channel that possesses a large audience. Health educators can use the media channel to disseminate consumer health information to a large audience. However, it might be a challenge to initiate and keep active discussion about health information among participants. Thus, health educators need to constantly feed health information to the online community.

In addition, physical activities, though frequently occurring in the discussion, was also negatively associated with conversational interactivity. A possible explanation is that physical activities are highly individualized. For example, some participants did crunches regularly, and others preferred water aerobics or snowshoeing. As members often had diverse interests in various sports, it was comparatively difficult for a conversation to maintain interactivity on a specific type of physical activity.

It should be noted that encouragement, a type of efficacious content, was negatively associated with conversational interactivity. A possible explanation is that a post often encouraged people to perform certain behaviors, such as posting in the discussion thread. Participants responded to the encouragement by performing the behavior rather than commenting on encouragement. Thus, posts with encouragement were less likely to have high conversational interactivity.

Limitations and future research

While the relationship between content features and interactivity was examined by this study, the results cannot be used to make predictions about causal relationships between content features, conversational interactivity, and actual health outcomes. Future studies should examine the connection between interactive conversation and the actual influence on health outcomes.

Additionally, the sampling period may contain more relevant discussion about weight loss than other periods of time during the year. However, frequencies of message features might not be representative of the entire year. This study examined only one particular online community although it has a high frequency of activities. Thus, findings from this study may not be generalizable to other online weight loss communities or communities related to other conditions.

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.

References

Ballatine

Stephenson

(2011) Help me, I’m fat! Social support in online weight loss networks. Journal of Consumer Behaviour 10(6): 332–337.

Bandura

(1991) Self-efficacy mechanism in physiological activation and health-promoting behavior. In: Madden

IV (ed) Neurobiology of Learning, Emotion and Affect, vol. 4. New York: Raven Press, pp. 229–270.

Bandura

(1993) Perceived self-efficacy in cognitive development and functioning. Educational Psychologist 28(2): 117–148.

Benigeri

Pluye

(2003) Shortcomings of health information on the Internet. Health Promotion International 18(4): 381–386.

Birru

Monaco

Charles

et al . (2004) Internet usage by low-literacy adults seeking health information: An observational analysis. Journal of Medical Internet Research 6(3): e25.

Brem

(2002) Analyzing online discussions: Ethics, data, and interpretation. Practical Assessment, Research and Evaluation 8(3). Available at: http://PAREonline.net/getvn.asp?v=8&n=3 (accessed 17 January 2014).

Bucy

Tao

(2007) The mediated moderation model of interactivity. Media Psychology 9(3): 647–672.

Burgoon

Bonito

Ramirez

Jr et al . (2002) Testing the interactivity principle: Effects of mediation, propinquity, and verbal and nonverbal modalities in interpersonal interaction. Journal of Communication 52(3): 657–677.

Centers for Disease Control and Prevention (2014) Healthy people: Nutrition and weight loss. Available at: http://www.healthypeople.gov/2020/ (accessed 1 October 2014).

10.

Chang

Chopra

Zhang

et al . (2013) The role of social media in online weight management: Systematic review. Journal of Medical Internet Research 15(11): e262.

11.

Cline

RJW

Haynes

(2001) Consumer health information seeking on the Internet: The state of the art. Health Education Research 16(6): 671–692.

12.

Cotton

Gupta

(2004) Characteristics of online and offline health information seekers and factors that discriminate between them. Social Science & Medicine 59(9): 1795–1806.

13.

De la Peña

Quintanilla

(2015) Share, like and achieve: The power of Facebook to reach health-related goals. International Journal of Consumer Studies 39(5): 495–505.

14.

Dutta-Bergman

(2004) Health attitudes, health cognitions, and health behaviors among Internet health information seekers: Population-based survey. Journal of Medical Internet Research 6(2): e15.

15.

Fox

(2013) Health and Technology in the U.S. Pew Research Internet Project. Available at: http://www.pewinternet.org/2013/12/04/health-and-technology-in-the-u-s/ (accessed 12 May 2014).

16.

Gil de Zúñiga

Jung

Valenzuela

(2012) Social media use for news and individuals’ social capital, civic engagement and political participation. Journal of Computer-Mediated Communication 17(3): 319–336.

17.

Ginossar

Nelson

(2010) Reducing the health and digital divides: A model for using community-based participatory research approach to E-health interventions in low-income Hispanic communities. Journal of Computer-Mediated Communication 15(4): 530–551.

18.

Gold

Burke

Pintauro

et al . (2007) Weight loss on the web: A pilot study comparing a structured behavioral intervention to a commercial program. Obesity 15(1): 155–164.

19.

Harvey-Berino

Pintauro

Buzzell

et al . (2004) Effect of Internet support on the long-term maintenance of weight loss. Obesity Research 12(2): 320–329.

20.

Hayes

Krippendorff

(2007) Answering the call for a standard reliability measure for coding data. Communication Methods and Measures 1: 77–89.

21.

Hwang

Ottenbacher

Green

et al . (2010) Social support in an Internet weight loss community. International Journal of Medical Informatics 79(1): 5–13.

22.

Jones

Fox

(2009) Generations online in 2009. Available at: http://pewresearch.org/pubs/1093/generations-online (accessed July 15 2014).

23.

Kaplan

Haenlein

(2010) Users of the world, unite! The challenges and opportunities of social media. Business Horizons 53: 59–68.

24.

Kiesler

Siegel

McGuire

(1984) Social pyschological aspects of computer-mediated communication. American Psychologist 39: 1123–1134.

25.

Liang

Scammon

(2011) E-word-of-mouth on health social networking sites: An opportunity for tailored health communication. Journal of Consumer Behaviour 10(6): 322–331.

26.

McMillan

Hoy

Kim

et al . (2008) A multifaceted tool for a complex phenomenon: Coding web-based interactivity as technologies for interaction evolve. Journal of Computer-Mediated Communication 13(4): 794–826.

27.

Marketdata (2013) Worldwide: U.S. weight loss market forecast to hit $66 billion in 2013. Available at: http://www.prweb.com/releases/2012/12/prweb10278281.htm (accessed 1 October 2013).

28.

Muntinga

Moorman

Smit

(2011) Introducing COBRAs: Exploring motivations for brand-related social media use. International Journal of Advertising 30(1): 13–46.

29.

Napolitano

Hayes

Bennett

et al . (2013) Using Facebook and text messaging to deliver a weight loss program to college students. Obesity 21(1): 25–31.

30.

Neuhauser

Kreps

(2010) eHealth communication and behavior change: Promise and performance. Social Semiotics 20(1): 9–27.

31.

Patrick

Koss

(1995) Consumer health information “white paper.” Working draft, Consumer Health Information Subgroup, Health Information and Application Working Group, Committee on Applications and Technology, Information Infrastructure Task Force, Washington, DC, 15 May.

32.

Pew Internet & American Life Report (2014) Health fact sheet. Available at: http://www.pewinternet.org/fact-sheets/health-fact-sheet/ (accessed 28 July 2014).

33.

Rafaeli

(1988) Interactivity: From new media to communication. In: Hawkins

Wiemann

Pingree

(eds) Sage Annual Review of Communication Research: Advancing Communication Science, vol. 16. Beverley Hills, CA: SAGE, pp. 110–134.

34.

Rafaeli

Sudweeks

(1997) Networked interactivity. Journal of Computer-Mediated Communication 2(4). Available at: https://onlinelibrary-wiley-com.web.bisu.edu.cn/doi/10.1111/j.1083-6101.1997.tb00201.x/full (accessed 28 July 2014).

35.

Resnick

Janney

Buis

et al . (2010) Adding an online community to an Internet-mediated walking program. Part 2: Strategies for encouraging community participation. Journal of Medical Internet Research 12(4): e72.

36.

Rimer

Glanz

(2014) Theory at a Glance: A Guide of Health Promotion Practice. Bethesda, MD: National Institutes of Health. Available at: http://www.sneb.org/2014/Theory%20at%20a%20Glance.pdf (accessed July 28 2014).

37.

Rodgers

Chen

(2005) Internet community group participation: Psychosocial benefits for women with breast cancer. Journal of Computer-Mediated Communication 10(4). Available at: https://onlinelibrary-wiley-com.web.bisu.edu.cn/doi/10.1111/j.1083-6101.2005.tb00268.x/full

38.

Tardy

Hale

(1998) Getting “plugged in”: A network analysis of health-information seeking among “stay-at-home moms.” Communication Monographs 65(4): 336–357.

39.

Tate

Jackvony

Wing

(2003) Effects of Internet behavioral counseling on weight loss in adults at risk for type 2 diabetes. Journal of the American Medical Association 289(14): 1833–1836.

40.

Thackeray

Eggett

Neiger

et al . (2003) Measuring the relevance of evaluation criteria among health information seekers in the internet. Journal of Health Psychology 8: 71–82.

41.

Trochim

(2015) The Research Methods Knowledge Base (2nd edn). Available at: http://www.socialresearchmethods.net/kb/ (accessed 1 Jan 2015).

42.

Wang

Qiao

Peng

(2015) Is the size or the valence of proactive engagement associated with purchase intention? A case study of branded blogs of Starbucks. International Journal of Strategic Communication 9(3): 197–216.

43.

Wang

Willis

Rodgers

(2014) Consumer insights from peer-to-peer communication in an online health community of weight management. Journal of Consumer Health on the Internet 18(2): 143–156.

44.

White

Dorman

(2001) Receiving social support online: Implications for health education. Health Education Research 16(6): 693–707.

45.

Whitten

Smith

Munday

et al . (2008) Communication assessment of the most frequented breast cancer websites: Evaluation of design and theoretical criteria. Journal of Computer-Mediated Communication 13(4): 880–911.

46.

Willis

(2016) Patients’ self-efficacy within online health communities: Facilitating chronic disease self-management behaviors through peer education. Health Communication 31: 299–307.