Engagement of young adult cancer survivors within a Facebook-based physical activity intervention

Abstract

Few studies have examined how young adult cancer survivors use online social media. The objective of this study was to characterize Facebook engagement by young adult cancer survivors in the context of a physical activity (PA) intervention program. Young adult cancer survivors participated in one of two Facebook groups as part of a 12-week randomized trial of a PA intervention (FITNET) compared to a self-help comparison (SC) condition. A moderator actively prompted group discussions in the FITNET Facebook group, while social interaction was unprompted in the SC group. We examined factors related to engagement, differences in engagement by group format and types of Facebook posts, and the relationship between Facebook engagement and PA outcomes. There were no group differences in the number of Facebook comments posted over 12 weeks (FITNET, 153 vs. SC, 188 p = 0.85) or the proportion of participants that reported engaging within Facebook group discussions at least 1–2 days/week. The proportion of participants that made any posts decreased over time in both groups. SC participants were more likely than FITNET participants to agree that group discussions caused them to become physically active (p = 0.040) and that group members were supportive (p = 0.028). Participant-initiated posts elicited significantly more comments and likes than moderator-initiated posts. Responses posted on Facebook were significantly associated with light PA at 12 weeks (β = 11.77, t(85) = 1.996, p = 0.049) across groups. Engagement within Facebook groups was variable and may be associated with PA among young adult cancer survivors. Future research should explore how to promote sustained engagement in online social networks. ClinicalTrials.gov identifier: NCT01349153

INTRODUCTION

Young adult cancer survivors (YACS), diagnosed between the ages of 18 and 39, have unique psychosocial and medical needs that occur during the developmental transition into adulthood [1, 2]. Long-term and late effects of cancer treatment, including increased risks for cardiovascular disease, diabetes, obesity, recurrence, and second cancers, may present significant challenges to this population over the course of their lives [1, 3, 4]. Healthy lifestyle behaviors, such as regular physical activity (PA) and maintaining a healthy weight, have the potential to reduce these risks and improve quality of life in cancer survivors [5, 6]. However, about 60% of young adults diagnosed with cancer during adolescence or young adulthood are not meeting PA guidelines of 150 min/week of moderate-to-vigorous-intensity PA for cancer survivors [7], and the prevalence of obesity in this population is higher than among peers without a history of cancer (31 vs. 27%) [3]. Although PA interventions have demonstrated benefits for cancer survivors, including improvements in cardiovascular fitness [8], depression [9], fatigue [10], and quality of life [10, 11], few have focused on YACS [12–15].

eHealth interventions for PA have demonstrated effectiveness among a variety of populations [16–18], and social media and social networking sites (SNS) are emerging as potentially effective approaches for PA and weight-related interventions [19–22] and for cancer prevention and control [23]. SNS have the potential to broadly reach individuals and engage users in interactive discussions, and young adults are some of the most active users of sites such as Facebook and Twitter [24]. Use of social media is especially prevalent among YACS, who often turn to social media for health information, peer support, and to engage with other cancer survivors and numerous cancer-specific organizations focused on YACS [25, 26]. While YACS continue to seek online information about healthy behaviors [27], few studies have examined how SNS are used by young adult survivors [25]. Overall, there is a need to better understand how cancer survivors use social media [25, 28] and how to best utilize online social networks to promote positive behavior change [21, 29].

A growing number of published studies have used SNS to deliver interventions related to PA or weight loss [13, 20, 30–32]. Systematic reviews of studies to date have demonstrated modest evidence supporting the efficacy of these approaches in promoting PA or weight control [18–22] and concluded that the isolated effect of SNS on behavior change is largely unknown [19, 21, 22]. The few studies that examined engagement within the context of these SNS interventions have found positive associations between online social network use and health-related outcomes. Turner-McGrievy and Tate [33] found that increased engagement with Twitter (i.e., number of tweets) predicted greater weight loss in the context of a mobile-delivered weight loss intervention. In a randomized controlled trial of an Internet-delivered intervention to promote social support for PA, Cavallo et al. [34] demonstrated that PA outcomes in the online social networking arm were more strongly associated with individuals’ Facebook use than with the use of the structured Facebook intervention group. Engagement with a Facebook page used to deliver a weight loss intervention to college students was variable among participants and decreased over the first 21 months of the 2-year study [35]. Polls and pictures posted by the study’s health coach received the most responses from participants. Another study by Hales et al. [36] demonstrated that higher engagement with Facebook was associated with greater weight loss during the maintenance phase after an initial behavioral weight loss intervention. Additionally, this study identified the most engaging types of Facebook posts, finding that polls elicited the highest use and counselor-initiated posts prompted more engagement than participant-initiated posts [36]. Developing and prompting discussions to promote health and maintain continued engagement with SNS present challenges, and the role of the counselor or moderator in sustaining these discussions is not well understood [37]. Although several online social networks focused specifically on PA exist, little is known about how users engage with various features and outcomes related to engagement [38].

Among studies that have examined engagement of cancer survivors within Internet-delivered interventions or social media, most have focused on online support groups, message boards, or researcher-built interventions [39–41]. Characterization of active contributions, such as posts or messages to online support groups, has demonstrated low levels of engagement among cancer survivors, with participants posting less than one message per week on average over the course of 12- to 16-week interventions [42–44]. Few studies have examined the ways that cancer survivors engage within existing social media platforms, such as Facebook, that are commonly used by cancer survivors and the general population alike [26, 45]. Furthermore, little is known about how to use social media to facilitate support in the context of a behavioral intervention for cancer survivors. Overall, research is needed to examine how cancer survivors engage within online SNS [39], to elucidate how engagement with social media by cancer survivors may relate to health outcomes [46], and to better understand how social media can be used to promote healthy behaviors among cancer survivors [28].

Thus, the objective of this study was to describe how YACS engaged within a Facebook group in the context of a SNS-based PA intervention program. Understanding how YACS use SNS can inform the design and implementation of future interventions for this population. This study sought to answer the following research questions:

1. Do levels (objective and self-reported measures of interactions) and perceptions of participant engagement with a Facebook group differ by group format (moderator-led vs. participant-led)?

2. Do baseline characteristics predict level of Facebook engagement or differ among types of participant engagement (active, somewhat active, reader, inactive)?

3. Is there an association between level and types of participant engagement with a Facebook group and PA outcomes (weekly minutes of moderate-to-vigorous-intensity PA and light PA)?

4. Do levels of participant engagement with a Facebook group vary by type of prompt (category of prompt, moderator-initiated or participant-initiated post)?

METHODS

Participants

Data for this secondary analysis are from the Fostering Improvement Through Networking and Exercising Together (FITNET) study, a randomized trial of a SNS-based physical activity intervention for YACS. Detailed recruitment and study methods were described previously elsewhere [13]. Participants were recruited primarily through community-based cancer organizations that posted advertisements on various communication channels, including Facebook, Twitter, listservs, and e-mail. Recruitment advertisements were also distributed through a university mass communication e-mail system, flyers in hospital clinics, and cancer organization message boards. Participants (n = 86) met the following inclusion criteria: ages 21–39 at study entry; diagnosed with cancer (excluding non-melanoma skin cancer) at age 18 or older; ≥ 1 year beyond date of diagnosis; completed cancer treatment; English-speaking; no medical condition(s) that precluded unsupervised exercise; not currently engaging in at least 150 weekly minutes of moderate-intensity exercise; and had Internet access and an active Facebook account. All participants provided informed consent, and study procedures were approved by the institutional review board of the participating university.

Procedures and intervention conditions

Details and primary outcomes of the intervention study are presented elsewhere [13]. The FITNET intervention aimed to increase moderate-to-vigorous-intensity PA (MVPA) levels among YACS to 150 weekly minutes, the guidelines for cancer survivors [47]. After completing a baseline online questionnaire, participants were randomized to one of two groups: FITNET intervention or self-help comparison (SC) (Table 1). FITNET participants received four main components, based on social cognitive theory [48] constructs: (1) 12 weekly Facebook messages with enhanced, theory-based behavioral lessons on PA after cancer and links to publicly available websites related to PA and/or cancer survivorship; (2) access to a secret Facebook group, where the study moderator posted discussion prompts to promote interaction (e.g., questions, videos, electronic resources); (3) access to a goal-setting and self-monitoring website, which had a weekly goal-setting feature, exercise tracker, and graphs with feedback on progress; and (4) a pedometer to self-monitor steps. The Facebook messages received by FITNET participants offered behavioral strategies and educational lessons on PA (e.g., goal-setting, overcoming barriers, enlisting social support). Examples of the various posts the study moderator made to the FITNET Facebook group wall are presented in Table 2. The SC group was designed to equate message contact time, attention, and pedometers with the intervention group. Participants in this group received the following: (1) 12 weekly Facebook messages with general exercise information and website links, (2) access to a secret Facebook group without prompts for interaction, and (3) a pedometer. The Facebook messages received by SC participants included the same publicly available websites given to FITNET participants and general tips on PA (e.g., benefits of PA). The main differences between the groups were in the lesson content, type of discussion within the Facebook group, and access to the self-monitoring website (Table 1). Communications with study participants took place online, and participants completed online questionnaires at baseline and after 12 weeks. Participants received a $30 gift card as an incentive after completion of the follow-up online questionnaire. As previously reported, there were no group differences in self-reported MVPA over 12 weeks, while the FITNET group increased light activity by 135 min compared with the SC group (FITNET, 164 min/week vs. SC, 29 min/week; p = 0.032) [13].

Table 1.

Description of FITNET intervention and self-help comparison group components

Delivery mode	FITNET intervention	Self-help comparison
In-person	Pedometer	Pedometer
Facebook	12 weekly Facebook messages	12 weekly Facebook messages
	•Links to publicly available websites	•Links to publicly available websites
	•Lessons with guidance on physical activity and behavioral strategies (e.g., self-monitoring, enlisting social support)	•Basic information and tips on physical activity
	Facebook group with moderated discussion prompts	Facebook group without discussion prompts
Website	Self-monitoring website with goal-setting tool, exercise tracker, and personalized feedback graphs

Table 2.

Description of moderator-initiated and participant-initiated post types from both study groups

Types of initiated posts	Definition	Example
Physical activity discussion questions	Moderator post asking participants a question about their experience with physical activity	“What do you see as barriers to becoming physically active? What activities did you do in the past? Why did you stop?”
Cancer-related discussion questions	Moderator post asking participants a question about their cancer experience	“How would you describe your experiences post treatment? Do you consider yourself to be a cancer survivor or something else?”
Physical activity information	Moderator or participant post sharing resource and/or tips about physical activity	“Is cold weather impacting your workouts? Here’s a short video with tips on cold weather workouts (link to website).”
News, resources related to cancer survivorship	Moderator or participant post sharing news item, resource, or current event related to cancer	“The National Cancer Institute recently launched a new site featuring some cancer survivor stories (link to website).”
Reminder (Thursday)	Moderator post with weekly reminder for participants to set an exercise goal, log daily activity, check the Facebook group, and practice an activity related to the weekly lesson	“Before week 6 starts on Sunday, just a reminder to: 1. Complete your daily activity log for this week. 2. On Sunday, complete your goal setting for week 6. 3. Check out the Facebook group and chat with other members. 4. Try adding helpful social cues for activity.”
Study information, logistics	Moderator post related to study features and activities, including technical support information	“The FITNET website is back up and available for you to set your week 2 exercise goal and track your activity. Thank you for your patience!”
Introduction	Participant post that introduces themselves and shares experiences with cancer and/or activity	“Hello everyone! I’m a Stage XXX survivor--been in remission for 3 years! Really hoping this group helps motivate me to make a lifestyle change! It’s been hard to get back exercising (especially to get rid of my recent baby weight)!”
Participant question about study	Participant post asking about study activities or logistics	“How do we now track our activity? I went to the site and see the goalsetting and activity tracker bars but don’t see where I can input activity.”
Participant question about cancer, treatment	Participant post that queries other participants about cancer or treatment	“I am wondering if anyone else has had surgery to remove scar tissue and what has been your experience?”
Check-in	Participant post that asks about group progress or experience. May include statement about individual’s exercise progress	“how is it going for everyone? my arms are so sore - i added weights to my workouts.”
Physical activity accomplishment	Participant post about physical activity accomplishment(s) or milestone(s) achieved in the past or present	“I walked in a 5K Cancer Awareness walk with my family on Saturday!...”
Physical activity goal	Participant post about a physical activity goal they would like to achieve	“On Oct. 22 I will be walking a 5k and then at the 1st of nov. running a 5k...wish me luck...lol”

Measures

Demographics and health-related variables

At baseline, participants were asked to report their age, gender, race/ethnicity, marital status, education, employment status, whether living alone or with a child in the home, cancer type and treatment, date of diagnosis, height, and weight. Body mass index was calculated as weight (kg)/height × height (m²). Physical activity was measured using the Godin Leisure-Time Exercise Questionnaire [49, 50], a previously validated measure commonly used in studies of PA among cancer survivors [51–53].

Facebook engagement

Table 3 describes the objective and subjective measures of interactions within the Facebook secret groups. Facebook content (posts, comments, videos, website links, etc.) was collected from both groups for each participant on a daily basis by a doctoral student in public health, who served as the study coordinator.

Table 3.

Measures of Facebook engagement

Measure	Description of objective interaction with Facebook	Description of self-reported interaction with Facebook “Over the last 12 weeks, how often did you do the following”: (1 = less often or never to 6 = several times a day)
Response to moderator-initiated posts	Participant interactions with Facebook posts by study moderator. Measured with the number of comments in response to a post by the moderator	Post responses to questions the study coordinator posted on the Facebook group wall (FITNET only)
Response to participant-initiated post	Participant interactions with Facebook post initiated by peer. Measured with the number of comments in response to participant-initiated post	Post responses to questions other participants posted on the Facebook group wall (SC only)
Likes	Participant interacts with either moderator or peer posts by clicking on “Like” icon next to post. Measured with the number of likes for original post, or any comments on the post. Likes for initiated posts are tallied only from original posts, not including likes on comments related to the original post	Click the “Like” button next to other people’s comments on the Facebook group wall
Participant-initiated posts	Participant initiates interaction with Facebook group by posting comment, question, or resource. Measured with the number of posts not in response to a discussion question or prompt posted by the moderator	Post a status, comments, questions, or information to the Facebook group wall
Read Facebook group discussion		Read Facebook group discussion

Objective measures

of Facebook engagement for each participant were summed over the course of the study and defined at the post-level as follows. Response posts were those made by a participant in response to or interacting with a post initiated by either the study moderator or a peer. Posts comprised either commenting or answering moderator prompts (responses to moderator-initiated posts) or commenting on posts by peers (responses to participant-initiated posts). Likes, or the number of times that a participant clicked the like button next to a post, were summed for each participant. Participant-initiated posts were the number of posts made by participants to the Facebook group that were independent of the moderator prompting discussion. Interactions for each participant were calculated from the sum of response posts, initiated posts, and likes. All objective measures were summed at the end of the study, based on the data collected every day, and did not account for whether a post or like was subsequently deleted or “unliked” after initial daily data collection.

Self-report measures

Prior Facebook use was assessed at baseline with a single item from the Facebook Intensity Scale (0 = less than 10 min to 5 = more than 3 h) [54]. In the post-intervention questionnaire, all participants were asked to report how often they engaged in four different interactions with Facebook, with response options ranging from 1 = less often or never to 6 = several times a day. Self -reported Facebook engagement questions asked both FITNET and SC group participants, “Over the last twelve weeks, how often did you do the following?: read Facebook group discussions; post responses to questions the study coordinator (SC: other participants) posted; post a status, comments, questions or information to the Facebook group wall; click ‘like’ button next to other people’s comments on the Facebook group wall.” Participants with missing responses at 12 weeks (n = 20) were coded as never engaging. At 12 weeks, participants were asked if they recalled any questions posted by the study coordinator (SC: other participants) on the Facebook group wall to prompt group discussion. Study completers who answered “yes” (n = 52, 60.5% of sample) were asked six additional questions about the group discussions, with responses ranging from 1 = not at all to 5 = completely; those who answered “no” were coded as 1 = not at all. Questions on perceptions of Facebook group discussions asked whether participants found information in Facebook group discussions to be designed especially for me and my needs; important to me personally; applies to my life; caused me to become physically active; motivating; and trust that information was accurate. Perceptions of Facebook group members were assessed at 12 weeks by asking participants how much they agreed with statements about members of their Facebook group (i.e., were motivating, were supportive). Responses ranged from 1 = strongly disagree to 5 = strongly agree.

Type of engagement

We further characterized participants by degree of engagement, based on both objective (posts, likes) and self-report (read group discussions) measures, categories used in previous studies [33–35], and a dichotomous cutpoint (median total Facebook interactions). Active participants were defined as those who interacted with Facebook at least twice over the course of the study and reported reading group discussions at least “1–2 days per week.” Participants who interacted with Facebook at least twice but reported rarely reading discussions were classified as Somewhat Active. Those who reported reading group discussions but interacted less than twice were categorized as Readers, while those who neither posted regularly, nor read discussions regularly, were considered Inactive.

Types of initiated posts

All posts initiated by the moderator or participants were coded by content into one mutually exclusive category, as described in Table 2. Six types of moderator-initiated posts were designed a priori to promote group discussion and elicit peer support within the FITNET group. In month 1, the moderator posted two discussion questions per week and some introductory information about study logistics (e.g., Facebook privacy, self-monitoring website). In months 2 and 3, discussion questions were tapered to one per week. To categorize participant-initiated posts, three coders (doctoral students in public health) reviewed the text of all Facebook posts and developed codes for common themes that emerged from participants’ posts. Each participant-initiated post was then categorized, based on codes applied by at least two of the three coders, as one of eight post types described in Table 2. For post-level analyses, we tallied the numbers of related comments and likes in response to each initiated post.

Statistical analysis

To characterize Facebook engagement, we calculated descriptive statistics for participant posts and likes and compared study groups on these and self-reported Facebook interactions. Differences in levels of engagement between the moderator-led and participant-led group formats were assessed using chi-square or Fisher’s exact tests for categorical variables or Mann-Whitney U tests for continuous variables, due to the skewed nature of the data. Related samples Wilcoxon signed rank tests were used to examine change over time in the number of participant interactions with Facebook. Spearman correlations were conducted to examine the association between objective and self-reported measures of Facebook engagement. Next, linear regression was used to examine predictors of levels of Facebook engagement, including baseline demographics, PA, and prior Facebook use. We examined baseline characteristics among the four types of Facebook engagement by conducting chi-square analyses for categorical variables or ANOVA with Tukey’s HSD for post hoc analyses of continuous variables. To address the third research question, we explored whether engagement with Facebook was associated with PA outcomes after the 12-week study across study groups. We examined residualized change in PA among the four categories of Facebook engagement by conducting Kruskal-Wallis tests with Dunn-Bonferroni adjustment and performed linear regression analyses, controlling for baseline activity and age. Analyses were intent-to-treat, including all participants, with baseline observations carried forward for any missing post-intervention data (n = 20). Finally, we analyzed post-level data by examining differences in Facebook engagement (i.e., number of comments and likes) among the types of initiated posts using Kruskal-Wallis tests and post hoc pairwise comparisons with Dunn-Bonferroni adjustment. We used Mann-Whitney U tests to compare the number of comments and likes between moderator-initiated and participant-initiated posts. All analyses were conducted using SPSS 23.0 (SPSS Inc., Chicago, IL).

RESULTS

Baseline characteristics of study participants are described in Table 4. Eighty-six YACS were randomized to participate in the 12-week intervention trial. A total of 66 (77%) participants completed post-intervention assessments with no difference in retention between study arms. Participants were mostly female (91%), non-Hispanic white (91%), and survivors of hematologic cancer (31%) with a mean age of 31.7 ± 6.2 years. On average, participants were overweight and almost 5 years beyond diagnosis. Study groups were comparable on baseline characteristics, with the exception of daily Facebook use, which was slightly higher in the FITNET group.

Table 4.

Baseline demographic characteristics by study group

Characteristics	FITNET Intervention (n = 45)	Self-help comparison (n = 41)	p value
Age (years), (mean + SD)	30.8 (5.7)	32.7 (4.2)	0.078
Sex [n (%)]
Female	41 (91.1)	37 (90.2)	1.000
Male	4 (8.9)	9 (0.8)
Race/ethnicity [n (%)]			0.470
Non-Hispanic white	42 (93.3)	36 (87.8)
Marital status [n (%)]
Married or living as married	21 (46.7)	22 (53.7)	0.517
Not married	24 (53.3)	19 (46.3)
Education [n (%)]
Less than college degree	9 (20.0)	10 (24.4)	0.624
≥College graduate	36 (80.0)	31 (75.6)
Annual income [n (%)]			0.960
<$50,000	20 (44.4)	18 (43.9)
≥$50,000	25 (55.6)	23 (56.1)
Employed full-time, %	20 (44.4)	20 (48.8)	0.687
Living alone	7 (15.6)	7 (17.1)	0.849
Responsible for children <18 years	17 (37.8)	16 (39.0)	0.905
Months since diagnosis	63.2 (52.1)	52.7 (32.7)	0.261
Body mass index (kg/m2), mean (SD)	28.4 (8.2)	29.1 (8.9)	0.721
≥3 h daily Internet use, %	25 (55.6)	27 (65.9)	0.329
Daily Facebook use,a mean (SD)	2.6 (1.4)	2.0 (1.0)	0.049
MVPA (min/week), mean (SD)	109.3 (125.0)	118.4 (126.3)	0.739
Mild PA (min/week), mean (SD)	78.3 (91.8)	81.0 (78.5)	0.887

MVPA moderate-to-vigorous-intensity physical activity, PA physical activity

^a0 = less than 10 min, 1 = 10–30 min, 2 = 31–60 min, 3 = 1–2 h, 4 = 2–3 h, 5 = more than 3 h

Do levels and perceptions of Facebook engagement differ by study group format?

Facebook engagement by study groups is shown in Table 5. Over the course of the 12-week study, participants engaged in 443 total Facebook interactions, including 341 posts and 102 likes. FITNET participants posted a total of 153 comments to the Facebook group compared to 188 comments by SC participants in the participant-led group. Over two thirds of participants posted at least once during the study (67.4%; n = 58/86; FITNET, 71.1%; SC, 63.4%), and over one third liked at least one post (34.9%; n = 30/86; FITNET, 28.9%; SC, 41.5%). Among those that posted or liked a post, median posts per person was 4 (1, 6.3) (range 1–36), while median likes per person was 2 (1, 3) (range 1–26). There were no differences between study groups in the median number of response posts or likes by participants. However, the SC group had significantly more participant-initiated posts than the FITNET group (p = <0.0005). The proportion of participants that made any interactions decreased over time in both groups; over half interacted in the first month (FITNET, 71.1% vs. SC, 63.4%; χ ²(1, n = 86) = 0.579, p = 0.447), and few posted in month 2 (11.1 vs. 12.2%; p = 1.000) or month 3 (8.9 vs. 7.3%; p = 1.000). Across both groups, the number of interactions was significantly lower in month 2 (Z = 31.00, p = <0.0005) and month 3 (Z = 0.00, p = <0.0005) compared to month 1 (Fig. 1).

Table 5.

Interactions with Facebook and types of participant engagement by study group (n = 86)

Variable	Total (n = 86)	FITNET intervention (n = 45)	Self-help comparison (n = 41)	p value
Objective engagement
Number of interactions	443	204	239
Median (IQR) per participant	2 (0, 7)	2 (0, 5.5)	2 (0, 7)	0.947
Mean (SD) per participant	5.15 (8.50)	4.53 (7.33)	5.83 (9.68)
Number of response posts	283	147	136
Median (IQR) per participant	1 (0, 4)	2 (0, 5)	0 (0, 3.5)	0.120
Mean (SD) per participant	3.29 (5.65)	3.27 (4.43)	3.32 (6.80)
Number of participant-initiated posts	58	6	52
Median (IQR) per participant	0 (0, 1)	0 (0, 0)	1 (0, 2)	<0.0005
Mean per participant	0.67 (1.32)	0.13 (0.51)	1.27 (1.66)
Number of likes	102	51	51
Median (IQR) per participant	0 (0, 1)	0 (0, 1)	0 (0, 1)	0.262
Mean per participant	1.19 (3.38)	1.13 (3.97)	1.24 (2.63)
Self-reported engagement
Number reporting engaging 1–2 days/week or more, n (%)
Read group discussion	31 (36.0)	14 (31.1)	17 (41.5)	0.318
Post responses to group wall	11 (12.8)	3 (6.7)	8 (19.5)	0.075
Post status, comments, questions	7 (8.1)	3 (6.7)	4 (9.8)	0.704
Like other people’s comments	10 (11.6)	3 (6.7)	7 (17.1)	0.183
Perceptions of Facebook group discussionsa
Number reporting somewhat-completely, n (%)
Designed especially for me	33 (50)	14 (43.8)	19 (55.9)	0.323
Important to me personally	34 (51.5)	16 (50.0)	18 (52.9)	0.811
Applies to my life	35 (53.0)	16 (50.0)	19 (55.9)	0.632
Caused me to become active	27 (40.9)	9 (28.1)	18 (52.9)	0.040
Motivating	32 (48.5)	13 (40.6)	19 (55.9)	0.215
Trust information was accurate	35 (53.0)	18 (56.3)	17 (50.0)	0.611
Perceptions of Facebook group membersa
Number reporting agree-strongly agree, n (%)
Members were motivating	25 (37.9)	10 (31.3)	15 (44.1)	0.281
Members were supportive	36 (54.5)	13 (40.6)	23 (67.6)	0.028
Types of participant engagementb
Number of participants, n (%)				0.624
Active	22 (25.9)	11 (24.4)	11 (26.8)
Somewhat active	22 (25.9)	12 (26.7)	10 (24.4)
Reader	9 (10.5)	3 (6.7)	6 (14.6)
Inactive	33 (38.4)	19 (42.2)	14 (34.1)

IQR interquartile range, SD standard deviation

^aSample includes study completers (n = 66: FITNET, n = 32; SC, n = 34)

^bActive: ≥ 2 interactions with FITNET group and reported reading group discussions at least 1–2 days/week; Somewhat active: ≥ 2 interactions with FITNET group and read group discussions <1–2 days/week; Reader: <2 interactions with FITNET group and read group discussions at least 1–2 days/week; Inactive: <2 interactions with FITNET group and read group discussions <1–2 days/week

Fig. 1.

Number of Facebook posts and likes by study group over time. Note: Posts include both response and participant-initiated posts. FITNET FITNET intervention group, SC self-help comparison group

Results of intent-to-treat analyses showed that the proportion of participants that reported engaging in Facebook activities 1–2 days/week was comparable between study groups (Table 5). Over a third of all participants read group discussions, while over 10% reported posting a response to questions or liking other people’s comments. Fewer than 10% of participants reported initiating posts with comments, questions, or information. Objective measures of Facebook engagement were significantly correlated with self-reported measures in both groups (response posts: r _s = 0.443; p < 0.0005; initiated posts: r _s = 0.364; p = 0.001; likes: r _s = 0.429; p < 0.0005).

Among study completers, both groups had comparable perceptions of information in the Facebook group discussions (Table 5). Most SC group participants had favorable perceptions, reporting that the information was somewhat-completely designed especially for me, important to me personally, applies to my life, motivating, and caused them to become physically active. Compared to the FITNET group, significantly more SC group participants endorsed that information in the group discussions caused them to become physically active (FITNET, 28.1% vs. SC, 52.9%, p = 0.040) and found members of the FITNET group to be supportive (FITNET, 40.6% vs. SC, 67.6%, p = 0.028).

Do baseline characteristics predict level of Facebook engagement or differ among reported types of participant engagement?

Baseline demographic characteristics were not associated with Facebook interactions (total posts and likes over the 12-week study), and neither was baseline Facebook use. Among study groups combined, over half of participants were Active or Somewhat Active with Facebook (n = 44, 51.2%), with another 10% (n = 9) reporting that they consistently read Facebook discussions. Baseline characteristics did not significantly differ among the four types of engagement groups (data not shown), and there were no differences between study groups in the proportion of participants categorized by type of engagement (Table 5). The proportion of participants that completed the 12-week study differed by type of engagement (χ ²(3, n = 66) = 18.5, p = <0.0005); all Active (n = 22, 100%) and Reader participants (n = 9, 100%), while only 77.3% of Somewhat Active (n = 17) and 54.5% of Inactive (n = 18) participants, completed the study.

Is there an association between level of Facebook engagement and physical activity changes at 12 weeks?

In exploratory analyses of the relationship between Facebook engagement and PA outcomes at 12 weeks, there were significant differences among the four engagement groups in change in weekly minutes of MVPA (χ ²(3, n = 86) = 11.323, p = 0.010). Pairwise comparisons showed that the Active group reported significantly greater change in MVPA than the Somewhat Active (Mdn = 18.5 vs. −48.4, U = 23.773, p = 0.009) and Inactive groups (Mdn = 18.5 vs. −46.5, U = 18.500, p = 0.042), but not the Reader group (Mdn = 18.5 vs. 11.6, U = 13.480, p = 1.000). Change in light PA over 12 weeks did not differ by engagement groups. In linear regression analyses adjusting for baseline light PA and age, response posts on Facebook significantly predicted light PA at 12 weeks, such that every increase of one response post in Facebook was associated with an increase of approximately 12 min/week in light-intensity PA (β = 11.77, t(85) = 1.996, p = 0.049).

Do levels of Facebook engagement differ by type of prompt or initiated posts?

There were 55 moderator-initiated and 58 participant-initiated posts during the study (Table 6). Among the six types of moderator-initiated posts, there were significant differences in response comments (p < 0.0005) and likes (p < 0.0005). Cancer-related discussion questions received the most responses from participants and were more engaging compared to reminder (p = 0.013) and study logistic (p = 0.035) posts. Moderator-initiated discussion questions about PA were the second most engaging and also received more comments relative to posts with reminders (p = 0.002) or study logistics (p = 0.009). Response comments on participant-initiated posts also significantly differed among the six categories of posts (p = 0.046), while likes did not. Compared to moderator-initiated posts, participant-initiated posts elicited significantly more response comments (p = 0.003) and likes (p = 0.002).

Table 6.

Number of comments and likes by moderator-initiated and participant-initiated post types from both study groups

Type of post	Number of posts	Number of comments in response to post Median (IQR) Mean (SD)	Number of likes in response to post Median (IQR) Mean (SD)
Moderator-initiated posts	55
Cancer-related discussion question	4	3.5 (1.3, 23.8)a 9.5 (13.8)	0 (0, 0) 0 (0)
Physical activity discussion question	13	2 (1, 4)b 3.5 (4.2)	0 (0, 0) 0.08 (0.3)
Physical activity resource/information	6	0.5 (0, 1.3) 0.7 (1.4)	1 (0, 1)c 0.7 (0.5)
Cancer-related news	8	0 (0, 0.8) 0.6 (1.4)	0 (0, 0) 0 (0)
Reminder (Thursday)	12	0 (0, 0) 0.2 (0.6)	0 (0, 0) 0 (0)
Study logistics, technical support information	12	0 (0, 0) 0.7 (2.0)	0 (0, 0) 0 (0)
p value for difference among post types		<0.0005	<0.0005
Participant-initiated posts	58
PA information/CA news	4	0.5 (0, 1.8) 0.8 (1.0)	0 (0, 1.5) 0.5 (1.0)
Introduction	20	1 (0, 2) 1.5 (2.4)	0 (0, 1) 0.5 (0.8)
Question about study logistics, technical support	9	3 (1, 4) 3.2 (2.2)	0 (0, 0) 0 (0)
Question about cancer, treatment	2	2.5 (2, 3) 2.5 (0.7)	0 (0, 0) 0 (0)
Check-in	10	1.5 (1, 5) 3.7 (4.7)	0 (0, 1) 0.4 (.5)
Physical activity accomplishment	9	3 (1, 8.5) 5.2 (6.5)	1 (0, 2.5) 1.3 (1.5)
Physical activity goal	4	1 (0.3, 1.8) 1 (0.8)	0.5 (0, 1.8) 0.8 (1.0)
p value for difference among post types		0.046	0.153
Total responses to moderator-initiated posts		0 (0, 2) 1.9 (4.7) Range 0–30	0 (0, 0) 0.1 (0.3) Range 0–1
Total responses to participant-initiated posts		1 (0, 3.3) 2.7 (3.8) Range 0–19	0 (0, 1) 0.5 (0.9) Range 0–4
p value for moderator vs. participant initiated		0.003	0.002

IQR interquartile range, SD standard deviation

^aSignificantly different from Reminder (p = 0.013) and Study logistics (p = 0.035) based on post hoc pairwise comparisons with Dunn-Bonferroni adjustment

^bSignificantly different from Reminder (p = 0.002) and Study logistics (p = 0.009) based on post hoc pairwise comparisons with Dunn-Bonferroni adjustment

^cSignificantly different from five other groups (ps = 0.006 to <0.0005) based on post hoc pairwise comparisons with Dunn-Bonferroni adjustment

DISCUSSION

This study characterized engagement with Facebook by YACS participating in a PA intervention, showing that total number of interactions was comparable between moderator-led and participant-led discussions implemented in an existing popular online SNS. Objective measures of engagement (i.e., Facebook comments and likes) were variable among participants and declined over time in both study groups. Over half of participants posted at least two comments to the Facebook groups, with another 10% reporting that they regularly read Facebook group discussions. While self-reported engagement with Facebook was similar between groups, the SC group had more participant-initiated posts compared with the FITNET group, and SC participants were more likely to report that group discussions caused them to become physically active and that other group members were supportive. Furthermore, across both groups, participant-initiated posts garnered significantly more response comments and likes than moderator-initiated posts. Participants who were active within Facebook (i.e., ≥ 2 comments and regularly read group discussions) reported greater changes in MVPA over time compared with somewhat active and inactive participants. Overall, results from this study suggest that engagement within a Facebook group may be associated with PA in YACS, and participant-led discussions, in particular, hold potential for encouraging interaction and support for behavior change.

Given that the FITNET intervention was designed to encourage social support through moderator prompts to encourage interaction, our finding that overall Facebook interactions did not differ between groups was unexpected. The SC group represents the interactions that naturally emerged through participant-initiated posts without any prompting by a study moderator or counselor. Interestingly, the SC group had a comparable number of total posts over 12 weeks and significantly more participant-initiated posts relative to the FITNET group. Combined with the finding that participant-initiated posts elicited more responses than moderator-initiated posts, our results suggest that when peers initiated posts or discussion prompts, these posts may have been just as encouraging, or perhaps even more so, as when prompted by a study moderator. Whereas a participant-led Facebook group appeared to be more well received than a moderated one among this sample of YACS, a previous study found that moderated online support groups elicited more responses and reading of content than peer-led online support groups for breast cancer survivors [55]. Similarly, engagement with Facebook (i.e., comments, likes, views) was significantly higher for counselor-initiated posts compared with participant-initiated posts among participants that joined Facebook support groups after a weight loss study [36]. A possible explanation for the positive perceptions and higher self-initiated engagement in the SC group is a comfort level or experience among some participants who may have previously interacted with peer survivors in other popular Facebook groups, such as those hosted by non-profit cancer organizations. Indeed, YACS have specifically expressed interest in peer support in the context of health interventions [56] and indicated meeting peer survivors as an unmet need [57]. Thus, SC participants may have had greater motivation to connect with peers and respond to their prompts and could have perceived support to be more relevant when it came from a peer, rather than from a moderator.

We previously reported no difference in MVPA minutes per week between the FITNET and SC groups, with both groups improving on this primary outcome over time. In the current analyses, contrary to what we anticipated, SC participants were more likely to report that participation in the Facebook group caused them to become physically active and that they felt supported by group members. Although previous research has demonstrated that cancer survivors may derive benefits from peer-led online support communities, such as improvements in well-being, depression, empowerment, and hope [58–60], some have shown no improvements [61] or worse outcomes associated with participation in peer-led online communities [62] compared to controls. Most studies have been among breast cancer survivors. The current study provides evidence for the potential of peer-led online groups in SNS to encourage interaction and support for increased PA. It is unknown how SC participants would have engaged with additional components, such as the self-monitoring website; it is possible that the additional intervention strategies received by the FITNET group were overwhelming or distracted them from participation in the Facebook discussions. Since the number of posts was comparable between groups, it is also possible that reading moderator posts could have taken away from the time FITNET participants might have otherwise spent posting or responding to peers and perhaps suggests some threshold of engagement. Overall, future research is needed to compare the capacity of moderators and peers to elicit support for and influence positive health behavior change in the context of SNS and to isolate the effects of engagement within online social networking groups from other intervention components.

Over half of participants interacted with the Facebook group at least twice; the median level of engagement was higher than that of female undergraduate students that participated in a 12-week SNS intervention as part of a randomized trial to promote PA [34]. Among the 67 women that received an intervention that included a Facebook group, participants engaged in a median of approximately one interaction (i.e., post, like) within the group. In the current study, over two thirds of YACS posted at least one comment, similar to engagement by college students participating in weight loss interventions delivered via Facebook [32, 35].

In studies of online support groups in cancer survivors, engagement, as measured by mean posts or comments per person, has been higher compared with four posts over 12 weeks in the present study. In a moderated online support group for gynecologic cancer survivors with psychoeducational materials, participants posted an average of eight comments over 12 weeks [44], while breast cancer survivors posted 20 and 12 times over 12 weeks in moderated and peer-led online support groups, respectively [55]. Our study differs from these others among cancer survivors in that the Facebook group was only one of multiple intervention components over which participants may have divided their time and participation. Similar to many behavioral interventions using SNS, participant engagement with Facebook declined over time [21, 22, 33–35]. While there were no differences in demographic characteristics across the categories of engagement, there were differences in rates of study completion, with all Active and Reader participants completing the study. Previous studies among cancer survivors have also found that demographic characteristics have not differed between participants that actively participated in online support groups and those that were passive readers (i.e., lurkers) [55, 63]. Given that all of the Reader participants completed the study and that evidence shows that lurkers may experience some benefits, such as improved psychosocial well-being, from participation in online cancer communities [63], studies should examine how lurkers can benefit from engagement in popular SNS, as well as how lurkers can best be encouraged to improve PA and other health behaviors. Additionally, research is needed to identify strategies to increase and sustain engagement with behavioral interventions in the context of SNS that are commonly used in everyday life.

The participants who were most actively engaged with Facebook reported significantly higher changes in MVPA levels than those that were Somewhat Active or Inactive. Previous studies of SNS interventions have also found that higher engagement with Facebook or Twitter was associated with improvements in PA [34] or weight control [33, 36]. It is unclear why response posts within the Facebook group predicted increased light PA, but not MVPA. Given their exploratory nature, these results should be interpreted with caution, as some other unmeasured variable may explain variance in light PA. Taken together, our findings provide evidence to suggest that engagement with a Facebook group focused on providing support and encouragement may be associated with PA in YACS.

Engagement within the Facebook group varied by the type of moderator-initiated post, with discussion questions about cancer experiences and PA receiving the most responses from participants. Similar to other social media-based interventions for weight control, posts that prompted participants to provide responses were more engaging. Previous studies have shown poll votes and event invitations, Facebook features that enable specific responses, to be the most engaging among participants in weight-related interventions [32, 36]. In future studies using SNS groups to encourage peer support and interaction among cancer survivors, engagement may be improved by posting specific prompts, such as questions and discussion questions that invite responses from participants. Future research should also examine how to best involve cancer survivors and peers to encourage engagement in social networks, such as training peers to lead discussions or using a combination of moderator- and peer-led approaches. Studies that recruit or capitalize on existing social networks of cancer survivors to promote healthy behaviors may also be warranted.

This is the first study to examine the group format and types of posts that elicit Facebook engagement by YACS in the context of a PA intervention program. Strengths of this study include the use of both objective and self-report measures of engagement within an existing popular SNS. The study design allowed us to compare moderated and participant-led group formats, and the SC group received an active intervention through Facebook, allowing for a more robust test of the effects of the intervention components on PA outcomes. Although our study has several strengths, a number of limitations should be considered when interpreting results. At the time of the study, other objective measures of Facebook engagement were not available, including the number of times posts were viewed or shared. Participants may have spent time reading posts or received support from other Facebook friends outside of the study, which we were unable to objectively assess and account for in analyses. Because of the multicomponent interventions, we were unable to isolate the effects of Facebook engagement and peer support from other intervention components. Given the relatively small and homogeneous sample, findings may have limited generalizability to the diverse population of YACS. Study participants were limited to those with existing Facebook accounts and may have been more engaged on SNS than other YACS. Finally, there may have been other unmeasured factors, such as style of coping, which possibly moderated the association between engagement with Facebook and perceptions of group discussions and outcomes [60].

In summary, we found that a Facebook group was feasible for facilitating communication and support among YACS. Among the SC group, a minimal intervention and an unmoderated participant-led Facebook group was utilized, on topic, and well received. Overall, our findings suggest that peer-led Facebook group discussions have potential for promoting interactions among YACS. In the future, research should continue to evaluate ways to capitalize on SNS features and functionalities that promote engagement and support in peer-led groups. Further defining valid metrics of engagement, intervention adherence, and interactions on SNS are also important areas to advance the science of SNS-based behavioral interventions. Exploration of online social networking-based approaches that may facilitate social support, longer-term adherence, and positive behavior changes in larger samples is warranted.

Compliance with ethical standards

Disclosure

A portion of the findings in this manuscript was previously reported during conference presentations at the International Society of Behavioral Nutrition and Physical Activity 2013 Annual Meeting, the International Society for Research on Internet Interventions 2013 Scientific Meeting, and the 2014 Biennial Cancer Survivorship Research Conference.

Conflict of interest

The authors declare that they have no conflict of interest.

Ethical approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.

Statement on the welfare of animals

This article does not contain any studies with animals performed by any of the authors.

Informed consent

Informed consent was obtained from all individual participants included in the study.

Funding

This study was supported by a 2011 Society of Behavioral Medicine Distinguished Student Award for Excellence in Research, the UNC Lineberger Cancer Control Education Program (National Institutes of Health/National Cancer Institute funded; R25 CA057726), and the UNC Communications for Health Applications and Interventions Core (funded through the Gillings School of Global Public Health Nutrition Obesity Research Center (National Institutes of Health/National Institute of Diabetes and Digestive and Kidney Diseases funded; P30 DK56350) and the Lineberger Comprehensive Cancer Center (National Institutes of Health/National Cancer Institute funded; P30 CA16086)).