Incentivizing social support in the randomized mobile Lifestyle Intervention for Food and Exercise study: the impact of gamification on social support perceptions, provision, and receipt

Abstract

Background:

Few weight loss interventions have isolated the use of gamification through points to incentivize the social support needed for successful weight loss during a mobile intervention.

Purpose:

To compare a group that receives points [mLife (mobile Lifestyle Intervention for Food and Exercise)+points] vs. a group that did not receive points (mLife) for completing social support activities. Differences in social support perceptions, provision, and receipt were examined as part of a 12-month weight loss intervention. The mediating role of these activities with weight loss was also examined.

Methods:

Participants with overweight/obesity (n = 243 enrolled) were recruited to participate in a 12-month randomized trial delivered via the mLife app. Perceived social support was measured via survey [Multidimensional Scale of Perceived Social Support (MSPSS)] and by ratings of support received in the app. Social support provision was assessed via the number of social support activities completed in the mLife app over 12 months, and social support receipt was assessed as the number of social support activities a participant received over 12 months.

Results:

At 12 months, mLife+points participants had greater perceived social support (difference between groups: total MSPSS 0.52 ± 0.16, P < .01; ratings of support via app: 0.17 ± 0.07, P = .02) and greater rate ratio (RR) of providing social support (RR = 2.23, P < .001) and receiving social support (RR = 2.4, P < .001) as compared to mLife participants. Both social support receipt and provision mediated the relationship between group and 12-month weight loss.

Conclusions:

Incentivizing social support provision during a weight loss intervention via points is a potential way to increase social support perceptions, provision, and receipt.

Clinical Trial information:

The Clinical Trials Registration #NCT05176847.

Graphical abstract

Lay Summary

Participants who were randomized to a mobile weight loss intervention incentivizing them to provide social support to one another using points reported greater perceived social support and provided and received more social support than a nonpoints condition.

Introduction

The treatment of overweight and obesity is multifactorial and typically involves recommendations to reduce energy intake and increase physical activity as well as receive support for behavior change [1]. In addition, group-based weight loss treatment is often more effective than individual treatment [2], partly due the social support provided in a group treatment setting [3].

There has been an increase in behavioral interventions for weight loss being delivered entirely remotely, partly driven by the coronavirus disease of 2019 (COVID-19) pandemic [4]. Electronic or mobile health (eHealth/mHealth) interventions for weight loss can provide easy access to tools for diet, physical activity, and weight self-monitoring as well as delivery of intervention content and goal setting with feedback [4, 5]. Replicating the group support that is provided in traditional face-to-face behavioral weight loss treatment has been more difficult. Studies have utilized a wide range of methods to help participants engage in social support with one another, such as social media groups [6], text messaging [7], and group chat [8].

Despite these efforts, maintaining engagement with social support features in mHealth/eHealth interventions has been challenging with many studies showing a decline in engagement over time [9]. One potential way to encourage engagement with social support features in an mHealth intervention is to leverage gamification. Gamification, which uses elements of game design (points, badges, leaderboards) in nongame settings, has been used in public health interventions, including those that target diet and physical activity [10]. In particular, social gamification may focus on things like financial incentives to gamify an intervention or facilitate support [11]. While previous weight loss interventions have used financial incentives with varying degrees of success [2, 12], these incentives can be costly. Provision of points for behaviors is a scalable and low-cost way to provide incentives. However, few weight loss interventions have isolated the use of points to incentivize the social support provision needed for successful weight loss [13].

The mobile Lifestyle Intervention for Food and Exercise (mLife) study, which was previously described in detail [14, 15], was a 1-year randomized parallel behavioral weight loss intervention delivered via the mLife app. The intervention focused on gamifying provision of social support by providing participants with points for completing social support activities such as giving a thumbs up to a participant who self-monitored their diet for the day or reaching out to a participant who had not logged in for 24 h. While all participants earned points for completing social support activities, those randomized to the mLife+points group could see the points whereas those in the mLife group could not (i.e. only visible to the researchers).

The mLife study was informed by behavioral economics approaches to health, which focus on helping direct people toward an action by using strategies like incentivizing behaviors [16]. The study is also informed by Self-Determination Theory (SDT), which emphasizes the importance of autonomy, competence, and relatedness for psychological well-being and optimal motivation. Use of financial incentives is often seen as at odds with SDT in that financial incentives may lower intrinsic motivation to engage in healthy behaviors [17]. However, use of points can help to gamify an intervention, thereby potentially increasing the enjoyment (and intrinsic motivation) of completing the activities and can directly tie into the SDT constructs of relatedness, competence, and autonomy [13].

This paper examined whether incentivizing social support through points increased participants’ perceived social support, social support provision, and social support receipt, and tested whether these aspects of support mediated weight loss over the 12-month study. We hypothesized that there would be higher perceived social support as well as greater social support provision and receipt in the mLife+points group vs. the mLife group. Although the perceived social support questionnaire did not specify support from online or in-app networks, it did assess social support from friends, family, and significant others, which allowed for an examination of whether increases in social support in the mobile intervention can have spillover effects on feelings of social support from other sources, which has been found in other research [27]. We also hypothesized that each of these examined social support factors would mediate the relationship between participant group and weight loss.

Methods

The methods for the mLife study, which was conducted in two cohorts between 2022 and 2024, have been described in detail elsewhere [14, 15], including details on recruitment and inclusion/exclusion criteria. Briefly, adults (18–65 years, body mass index between 25 and 49.9 kg/m²) living in the United States, with >2 risk factors for type 2 diabetes [18], were recruited via social media ads and Researchmatch.org to participate in the 12-month study. The entire intervention, including assessments, was conducted remotely. Interested participants completed a screening questionnaire and phone call, and if they qualified, were invited to an orientation session delivered via Zoom where they learned more about the study and signed an informed consent document. Once all baseline assessments were completed, participants were invited to a training session specific to their randomized assignment and instructed on how to use the app. For body weight, participants were asked to step on their Fitbit scale first thing in the morning, in light clothing, at baseline, 6 months, and 12 months [14, 15]. The mLife study was overseen by a Data Safety Officer, and all research was approved by the University of South Carolina IRB.

Intervention overview

Participants received behavioral content via twice weekly podcasts and tips of the day that appeared on days without a podcast release. The mLife app had a newsfeed where participants could view and give a thumbs up to a group member when that member had self-monitored their diet for the day (1 point), logged at least 30 min of physical activity (1 point), stepped on the scale for a weight measurement (1 point), or listened to a podcast or read a tip of the day (1 point). In addition to providing a thumbs up for those activities, participants could also send an encouraging message to a group member who had not logged in for 24 h (1 point), post a request for help to the group chat on a specific topic related to diet or physical activity (1 point), respond to a request for help (1 point), and thank someone for a thumbs up or response to a request for help (1 point). Participants could also rate how supported they felt for each response they received. These activities resulted in a total of eight possible points earned per day (although participants could do more activities).

The only difference between the two groups was that the mLife+points group had the page detailing the social support tasks to complete each day for points (whereas the mLife group just had this presented as a daily checklist) and had a leaderboard, which ranked the number of points earned by each participant. This leaderboard was refreshed each month to sustain an element of achievable competition for all participants. Participants continued the intervention through 12 months and were provided with Amazon gift cards for completing assessments at both 6 and 12 months.

Survey measures of perceived social support

Most measures of social support were assessed objectively via the app. However, perceived social support was also assessed at baseline, 6 months, and 12 months via the Multidimensional Scale of Perceived Social Support (MSPSS) [19]. The MSPSS, which assesses perceived social support from friends, family, and a significant other, is a valid and reliable way to assess perceived social support [20]. The MSPSS has 12 items (four for each source of social support), and participants rated each statement from 1 (very strongly disagree) to 7 (very strongly agree) for a total of 4–28 for each social support source and a total for all 3 (total perceived support) of 12–84.

App measures of social support perceptions, provision, and receipt

When a participant received a response to a request for support on the discussion forum, they were able to thank that fellow participant for their support. Once they thanked them, the participant was asked to rate the level of support on a scale of one to five stars for each response (1 indicating a “poor support” and 5 indicating “excellent support”).

We previously reported the findings on the number of points (out of a possible 8 per day for 12 months) participants earned. While the mLife+points group could see their points, researchers were able to see the points that would have been earned from the mLife group as well. As hypothesized, the mLife+points participants earned more points (605.9 ± 203.2) than the mLife participants (350.0 ± 200.0; P < .01), indicating that social support provision was higher in the mLife+points group than the mLife group [15]. While participants could earn up to eight points per day for eight social support activities, participants were not limited in how many social support activities they did each day. Therefore, the present paper examines differences in the total number of activities and the number of activities within each category completed over 12 months.

The number of times participants received any type of support (e.g. thumbs up for completing a self-monitoring or intervention task, receipt of encouraging message to log onto the app, were thanked by a participant) was also assessed.

Statistical analysis

Complete power calculations have been presented elsewhere [14]. The study aimed to recruit 240 participants, which accounted for a predicted 20% attrition. Descriptive statistics were used to present baseline characteristics. The statisticians for the study were blinded to group assignment when completing all analyses.

Statistical analyses were conducted to analyze the differences in social support provision, receipt, and perception, as well as if those factors mediated weight loss. We did not observe particular patterns of participants’ missingness across all time points. Therefore, there was no evidence that the data were not missing at random. Missing data were addressed using maximum likelihood estimation in mixed modeling. Otherwise, complete case analysis was used. As we have applied the most appropriate statistical models in each data structure, no sensitivity analysis under a given set of scenarios was performed.

First, we adopted repeated measures mixed models to address how perceived social support (assessed via MSPSS) changed over time as well as what effect group assignment had. Mixed models were specified with main effects of group and time and interaction effect of group × time, adjusting for covariates of baseline education (≥college graduate or <college graduate), sex, and age. We then compared least square mean (LSM) differences in MSPSS from baseline to each assessment time point. These change values were contrasted across the groups at each time point for difference-in-difference analysis. PROC MIXED procedure in the SAS^® system (version 9.4) was used to fit mixed models with maximum likelihood estimation.

Second, analysis of covariance (ANCOVA) was performed to examine the effect of group assignment on perceived social support rating interactions at 12 months. Unlike MSPSS scores, which had ratings assessed at baseline, app measured social support was continuous and therefore calculated as a total number of activities at the end of the intervention. We estimated the LSM scores at 12 months for each group and contrasted them between the groups. The model included the group as a main effect, controlling for the same covariates used in mixed models. PROC GLM in the SAS^® system was processed.

Third, because of the large variance in social support activities across participants, Poisson regression models (vs. linear regression) were fit to the counts of social support activities completed over 12 months. For the several values that did not have a fixed upper bound, the counts of social support activities were assumed to follow the Poisson distribution. Using the same main effect and covariates included in ANCOVA model, we estimated the coefficient of the log count for number of social support activities associated with mLife+points group. To interpret the log of the change in the mean count given the treatment group assignment, rate ratios (RRs) were obtained by transforming the log coefficients with exponentiating. As a result, the exponentiated coefficients are reported as multiplicative increases or decreases in the social support activities for mLife+points group compared to mLife group. We used PROC GENMOD procedure with Poisson distribution and logit link function in the SAS^® system.

Lastly, the mediation effects of social support were investigated through MacKinnon’s product of coefficients test [21]. We estimated the intervention group effect on weight loss via the mediating social support variables. Potential mediating variables included changes in perceived social support (MSPSS) either at 6 months or at 12 months from baseline, app-based mean rating of provided social support, social support provision (total number of points), and social support receipt (total number of times participant received support). For changes in MSPSS, we controlled for baseline scores in the model. For the test of other mediators, we used scores measured at 12 months without controlling for earlier scores (since there were no baseline scores). We first estimated the effect of the intervention group on each mediator (a). We then estimated the effect of each mediator on weight loss at 12 months (b), controlling for intervention group effect and participants’ baseline weight. Finally, we calculated the product of coefficients by multiplying the a and b coefficients (ab). Because the indirect effects (i.e. product of a and b) rarely follow a normal distribution, we used bootstrapping to generate confidence intervals (CIs) for a, b, and ab. We obtained bias-adjusted empirical distributions of direct and indirect effects estimates with 5000 resamplings. If 95% CIs of the product of coefficients did not include zero, it indicated a significant mediator effect. The same covariates adjusted for the prior analyses were included in mediation analyses. Each individual mediation model was independently tested. The analyses were performed using lavaan R package v0.6.16 [22] and R Statistical Software v4.3.0 [23].

Results

The findings related to weight loss have been described elsewhere, including a CONSORT diagram [14, 15]. The participants in this study consisted of 243 adults with overweight/obesity (mean BMI 34.7 ± 6.0 kg/m²) and a mean age of 46.7 ± 11.6 years and a demographic composition of 67% White/22% Black, 87% female (see Supplementary Table 1 for complete demographics). A total of 199 participants completed a weight measure at 6 months (18% attrition), and 164 participants completed a weight measure at 12 months (33% attrition). Attrition was higher in the mLife group at 12 months (41%) than the mLife+points group (22%; χ² = 9.8, P < .01). There were no differences in baseline demographic characteristics between participants who completed a 12-month weight assessment and those who did not. Participants demonstrated a mean adherence (% days mLife app was used) of 40.9% (SD = 35.9) and a median (interquartile range [IQR]) of 27% (IQR: 0.27–100). Adherence was significantly greater in the mLife+points group (mean = 61%, median = 36%) compared with the mLife group (mean = 42%, median = 19%; χ² = 7.6, P < .01). For the perceived social support-related measures, 192 participants completed the MPSS at 6 months (21% attrition) and 161 completed at 12 months (34% attrition); and 149 participants provided at least one rating (1–5 stars) for support they received from other participants (39% did not provide a rating). For the number of participants who provided social support, 230 (95%) completed at least one social support activity. All participants received a social support activity from another participant (e.g. received a thumbs up for completing an activity) during the study with a mean of 2482.1 (mean of 6.8/day; range 9–10 980) received social support activities. One participant in the mLife+points group developed a thyroid disorder during the study and was removed from analyses [15]. Weight loss findings are provided elsewhere [15], and outcomes did not differ between the groups at 12 months (mLife+points −5.3 ± 0.6 kg and mLife participants −3.5 ± 0.7 kg; P = .09).

Perceived social support

Group by time interactions for total MSPSS scale 12 months, as well as the subscales (friends, family, and significant other) differed between groups, with the mLife+points group having greater improvements in total MSPSS and all subscales (Table 1). The mLife+points group also had greater improvements in the 6-month MSPSS family subscale.

Table 1.

Differences in perceived social support assessed via survey (MSPSS) and app (ratings of social support received on scale of 1–5) between groups at 6 and 12 months

	mLife+points group (n = 120)	mLife group (n = 122)	Difference between groups	Effect size for difference between groups (d)	P-value for difference between groups
MSPSS score analyzed via repeated measures models and presented as adjusted least squared mean ± SE (95% CI), effect size (d) a
Change in support from friends MSPSS subscale
6 months	0.27 ± 0.12 (0.04, 0.5) d = 0.16b	0.11 ± 0.13 (−0.14, 0.36) d = 0.07	0.17 ± 0.17 (−0.17, 0.51)	0.13	.34
12 months	0.45 ± 0.12 (0.21, 0.7) d = 0.27b	0.07 ± 0.14 (−0.19, 0.34) d = 0.05	0.38 ± 0.18 (0.02, 0.74)	0.31	.04
Change in support from family MSPSS subscale
6 months	0.17 ± 0.12 (−0.06, 0.4) d = 0.1	−0.23 ± 0.13 (−0.48, 0.03) d = −0.15	0.4 ± 0.18 (0.05, 0.74)	0.31	.02
12 months	0.28 ± 0.13 (0.03, 0.53) d = 0.16b	−0.24 ± 0.14 (−0.51, 0.03) d = −0.16	0.52 ± 0.19 (0.15, 0.89)	0.42	.01
Change in support from significant others MSPSS subscale
6 months	0 ± 0.14 (−0.27, 0.27) d = 0	−0.31 ± 0.15 (−0.6, −0.01) d = −0.21b	0.3 ± 0.2 (−0.09, 0.7)	0.20	.13
12 months	0.23 ± 0.15 (−0.06, 0.51) d = 0.14	−0.43 ± 0.16 (−0.74, −0.12) d = −0.3b	0.66 ± 0.21 (0.24, 1.08)	0.43	.002
Change in total MSPSS score
6 months	0.16 ± 0.1 (−0.05, 0.36) d = 0.11	−0.13 ± 0.11 (−0.35, 0.09) d = −0.1	0.29 ± 0.15 (−0.01, 0.59)	0.25	.06
12 months	0.33 ± 0.11 (0.11, 0.54) d = 0.23b	−0.19 ± 0.12 (−0.43, 0.04) d = −0.15	0.52 ± 0.16 (0.2, 0.84)	0.47	.0015
Perceived social support ratings of interactions in the app analyzed via ANCOVA and presented as adjusted least squared mean ± SE (95% CI) a
Mean rating of social support provided (1–5 stars with 5 being the highest rated)	4.83 ± 0.2 (4.44, 5.22)c	4.66 ± 0.19 (4.29, 5.04)c	0.17 ± 0.07 (0.03, 0.3)	0.34	.02

^aAll models adjusted for baseline sex, age, and education.

^bSignificant within-group changes, P < .05.

^cSignificant within-group differences, P < .05.

During the 12-month study, 149 (61%) participants rated at least one interaction (n = 81 in mLife+points and n = 68 in mLife; χ² = 3.22, P = .07), with more interactions receiving ratings (mean ± SE) in the mLife+points group (138.8 ± 25.7) than the mLife group (20.4 ± 4.2; P < .001). Of these rated interactions, mLife+points participants rated them higher (4.83 ± 0.2) than mLife participants (4.66 ± 0.19; P = .02; Table 1). Participants could also thank someone for providing support but then not rate the support, which resulted in a mean of 29.3 ± 2.8% of posts that were not rated. This did not differ between groups (28.7 ± 4.3% mLife+points group, 29.7 ± 3.7% mLife group; P = .85).

Social support provision

mLife+points participants were more likely to participate in all social support provision activities, as well as post or respond to the discussion board than mLife participants (Table 2). Overall, mLife+points participants engaged in social support provision activities 2.23 times more than mLife participants (95% CI: 0.80, 0.80; P < .0001). mLife+points participants also posted 6.07 times as many discussion board posts and response comments as the mLife group (95% CI: 1.77, 1.84; P < .0001).

Table 2.

Results of Poisson regression models examining the relationship between groups and social support provision measured within the app as the total number of social support activities completed, the number of activities by category, and total posts and comments to the discussion board^a

	Total number of social support activities completed over 12 months				Responded to a support request				Reached out to a nonresponsive participant				Provided thanks for receiving a thumbs up or comment
	RR	b	SE	P-value	RR	b	SE	P-value	RR	b	SE	P-value	RR	b	SE	P-value
Intercept	132.36	4.89	0.01	<.0001	0.63	−0.47	0.08	<.0001	5.34	1.68	0.07	<.0001	173.69	5.16	0.01	<.0001
mLife+points group	2.23	0.80	0.00	<.0001	5.54	1.71	0.02	<.0001	1.16	0.15	0.02	<.0001	1.96	0.67	0.00	<.0001
	Requested support from others				Provided a thumbs up for a participant who logged their weight				Provided a thumbs up for a participant who logged their diet				Provided a thumbs up for a participant who logged their physical activity
	RR	b	SE	P-value	RR	b	SE	P-value	RR	b	SE	P-value	RR	b	SE	P-value
Intercept	0.18	−1.73	0.18	<.0001	9.91	2.29	0.02	<.0001	26.37	3.27	0.02	<.0001	8.14	2.10	0.03	<.0001
mLife+points group	7.61	2.03	0.04	<.0001	2.00	0.69	0.01	<.0001	2.54	0.93	0.01	<.0001	1.99	0.69	0.01	<.0001
	Provided a thumbs up to a participant who read a tip of the day or listened to a podcast				Total number of posts to the discussion board				Total number of comments to the discussion board				Total number of both posts and comments
	RR	b	SE	P-value	RR	b	SE	P-value	RR	b	SE	P-value	RR	b	SE	P-value
Intercept	15.93	2.77	0.02	<.0001	0.22	−1.49	0.17	<.0001	0.86	−0.16	0.07	0.03	0.42	−0.86	0.07	<.0001
mLife+points group	1.71	0.53	0.01	<.0001	7.44	2.01	0.04	<.0001	5.06	1.62	0.02	<.0001	6.07	1.80	0.02	<.0001

^aAll models adjusted for baseline sex, age, and education.

Social support receipt

There was more social support received in the mLife+points groups as measured by the total times any support was received and within each social support activity as compared to the mLife group (Table 3). Overall, the mLife+points group was 2.4 times more likely to receive a social support activity by a fellow participant than the mLife group (95% CI: 0.87, 0.88; P < .0001).

Table 3.

Results of Poisson regression models examining the relationship between groups and social support receipt measured within the app as the total number of times a participant received a social support activity provided by another participant and the number of social support receipt activities by category^a

	Total number of times an individual received a social support activity by a fellow participant over 12 months				Received a response to their request for support				Received a message from another participant due to not logging into the app for 24 h				Received a thumbs up for logging their weight
	RR	b	SE	P-value	RR	b	SE	P-value	RR	b	SE	P-value	RR	b	SE	P-value
Intercept	447.02	6.10	0.01	<.0001	1.79	0.58	0.08	<.0001	49.02	3.89	0.05	<.0001	57.61	4.05	0.02	<.0001
mLife+points group	2.40	0.87	0.00	<.0001	5.31	1.67	0.02	<.0001	1.19	0.17	0.02	<.0001	2.30	0.83	0.01	<.0001
	Received a thumbs up for logging their diet				Received a thumbs up for logging their physical activity				Received a thumbs up for reading a tip of the day or listening to a podcast				Received a reply to a comment they left on the discussion board
	RR	b	SE	P-value	RR	b	SE	P-value	RR	b	SE	P-value	RR	b	SE	P-value
Intercept	183.17	5.21	0.01	<.0001	178.59	5.19	0.02	<.0001	84.06	4.43	0.02	<.0001	0.97	−0.03	0.23	0.90
mLife+points group	2.77	1.02	0.01	<.0001	2.17	0.78	0.01	<.0001	2.07	0.73	0.01	<.0001	2.30	0.83	0.04	<.0001

^aAll models adjusted for baseline sex, age, and education.

Social support mediation of weight loss

Supplementary Table 2 presents the mediation results testing whether the social support variables mediated the relationship between intervention group assignment and 12-month weight loss. The significant a paths indicate that participants in the mLife+points group showed greater increases in perceived social support changes at 12 months (P = .01), perceived social support via the app (P = .01), social support provision (P = .01), and social support receipt (P < .0001) than those in the mLife group. There was one significant b path; greater social support receipt was significantly associated with greater weight loss (P < .0001). Both social support provision (ab = −1.23, 95% CI = [−4.21, −0.10]) and social support receipt (ab = −8.90, 95% CI = [−13.96, −5.09]) were significant mediators in the pathway between intervention group assignment and weight loss. The indirect effect of X on Y through M (and we have two Ms) was significant, while the direct effect was not, consistent with full mediation. The mediation effect indicates that the mLife+points group’s greater weight loss at 12 months was due to greater provision (−1.23 kg) and receipt (−8.90 kg) of social support.

Perceived social support (both measured as MSPSS score and app-based mean ratings of provided social support) did not significantly mediate the relationship between intervention group assignment and weight loss.

Discussion

The mLife study examined the use of social gaming to incentivize participants to provide social support to one another over the course of a 12-month behavioral mobile health intervention. The present study found that all self-reported measures of social support (via the MSPSS) were higher in the mLife+points group (which was incentivized to provide social support) than in the mLife group (which was not incentivized) at 6 and 12 months. Furthermore, across all app-measured interactions of social support provision and receipt, mLife+points participants were more likely to complete social support activities and be the recipients of these activities than mLife participants. While none of the measures of perceived social support mediated the relationship between intervention and weight loss, objectively measured social support provision and receipt were mediators.

Other mHealth studies have also examined the role of social support in impacting outcomes. For example, a review found that social support features in technology-mediated interventions were important for improved diabetes self-management [24]. In addition, receiving social support in an mHealth group-based intervention was found to be associated with greater dietary self-monitoring and group interactions [25]. Not all studies have found that mHealth social support features impact health outcomes. In a systematic review of mHealth features in physical activity interventions, among four examined RCTs, having social features was not associated with physical activity outcomes and demonstrated that while some participants enjoyed the social competition or support, others did not find these features helpful [26]. mLife utilized leaderboards and not all participants may have found that helpful for weight loss, which may help to explain why perceived social support measures did not mediate weight loss.

In the present study, perceived social support was higher in the mLife+points group, as measured via survey and ratings of interactions in the app. While the MSPSS did not assess perceived social support specifically from the mLife app, it did assess perceptions from other sources. Previous research has found that feeling social support from one source can generalize to feeling of support from other sources [27]. For example, individuals receiving community-based support during a natural disaster also perceived receiving social support from friends and family [28]. Higher levels of perceived social support have been associated with better health outcomes, such as better stress response [29] and lower rates of depression [30]. Studies have also found among several populations that perceptions of social support are associated with both healthy eating and physical activity [31, 32]. Most of the differences in perceived social support occurred at 12 months rather than 6 months in the mLife study. A systematic review of social-support-based weight-loss interventions found that impacts on weight loss were only seen at the end of study or during follow-up periods [33], indicating that it may take time for perceived social support to increase and for social connected-based interventions to impact weight loss. In the present study, however, perceived social support was not a mediator for weight loss.

Social support provision was also higher in the mLife+points group than the mLife group. This variable was measured by both the eight types of activities participants could do to provide support (which earned points for mLife+points participants), as well as the number of posts and response comments made to the discussion board. The two activities that received the greatest RRs of social support provision in the mLife+points group were requesting support from other participants (RR = 7.61) and responding to support requests (RR = 5.54). Social support has been defined as “an exchange of resources between two individuals perceived by the provider or the recipient to be intended to enhance the well-being of the recipient” [34]. This exchange occurred much more frequently in the mLife+points group than the mLife group. While all the above-mentioned social support behaviors were incentivized with points in the mLife+points group, posting to the discussion board and responding to posts were not. However, the mLife+points group were almost twice as likely to interact on the discussion board than the mLife group, indicating that gamifying certain social support activities may translate over to other activities as well. Providing social support also mediated the relationship between intervention group and weight loss.

Lastly, mLife+points participants were 2.4 times more likely to receive a social support activity than the mLife group. Similarly to the findings of social support provision, the highest RRs of a receiving a social support activity was for receiving a response to a request for support. mLife+points participants were more than five times more likely to receive a response to their request for support than mLife participants. Online weight loss interventions can provide a unique platform to allow participants to receive social support [35]. Receiving social support during an online weight loss intervention may increase a participant’s sense of feeling seen by others and increase autonomy for weight loss-related behaviors, such as self-monitoring [36]. Receiving social support also was a significant mediator of weight loss. Both the mLife app and other online social support platforms for weight loss hold promise as ways for individuals to receive real-time, in-the-moment social support during weight loss [37].

The present study has several strengths. The study was able to isolate the impact of using points to gamify provision of social support in a weight loss intervention and have objective measures of social support activities over 12 months. A survey-based measure of social support was also used. The study also recruited participants from across the United States, broadening the generalizability of findings. The study also has limitations. Attrition was higher than expected, which may have hindered the ability to detect mediation effects for perceived social support. Participants were mostly white females, which limits applicability to males and more diverse populations. However, in a review of 94 weight loss interventions, average percent of participants who were African American was 18.2%, which is lower than the 22% in mLife [38]. Future mobile weight loss studies should examine ways to recruit and engage diverse populations. For example, weight loss interventions for African American adults may consider addressing healthy eating and physical activity barriers and providing culturally relevant content from healthcare professionals [39, 40]. In addition, African American men may prefer diet and exercise-related information for conditions other than just obesity, and African American women may prefer building community and having access to culturally relevant recipes [39]. Engaging men in weight loss research has also been difficult [41] and more women than men participate in organized weight loss programs [42]. Recruitment methods used in mLife, such as Researchmatch.org and social media, may have yielded a more technology-savvy participant sample than the general population. Lastly, the use of gamification may have targeted extrinsic vs. intrinsic motivation, which could prevent sustained behavior change.

Conclusion

The mLife study found that gamifying a mobile weight loss intervention through the provision of points was a simple way to increase social support perceptions, provision, and receipt; and that social support provision and receipt were both mediators of weight loss. Future studies may want to ensure that participants are incentivized to provide social support and also test if incentivizing and gamifying other areas of a weight loss intervention, such as dietary self-monitoring or physical activity tracking, may be associated with weight loss. In addition, future work may want to examine temporal trends in social support and how patterns may change over time. Lastly, additional mHealth studies should be conducted among more diverse samples.

Supplementary Material

Supplementary material is available at Translational Behavioral Medicine online.

Implications.

Practice:

Gamifying social support provision through the use of points in a mobile weight loss intervention may lead to great perceptions of social support and well as greater social support provision and receipt as compared to a nongamified approach.

Policy:

The use of points for incentivizing behaviors, such as social support provision, could be a low-cost way to help boost impacts of a mobile behavioral intervention.

Research:

Additional intervention research may wish to test if incentivizing and gamifying other areas of a weight loss intervention, such as dietary self-monitoring or physical activity tracking, may be associated with weight loss.

Data Availability

De-identified data from this study are not available in a public archive and will be made available (as allowable according to institutional IRB standards) by emailing the corresponding author.