Sharing digital self-monitoring data with others to enhance long-term weight loss: A randomized controlled trial

Abstract

Background:

Participants in behavioral weight loss (BWL) programs increasingly use digital tools to self-monitor weight, physical activity, and dietary intake. Data collected with these tools can be systematically shared with other parties in ways that might support behavior change.

Methods:

Adults age 18 to 70 with overweight/obesity (BMI 27–50 kg/m²) will enroll in a remotely delivered, 24-month BWL program designed to produce and maintain a 10% weight loss. Participants will be asked to use a wireless body weight scale, wearable activity sensor, and dietary intake app daily. All participants will receive individual and group counseling, engage in text messaging with members of their group, and appoint a friend or family member to serve in a support role. A 2×2×2 factorial design will test the effects of three types of data sharing partnerships: 1) Coach Share: The behavioral coach will regularly view digital self-monitoring data and address data observations. 2) Group Share: Participants will view each other’s self-monitoring data in small-group text messages. 3) Friend/Family Share: A friend or family member will view the participant’s data via automated message. The primary outcome is weight loss at 24 months. Mediators and moderators of intervention effects will be tested.

Conclusion:

This study will provide a clear indication of whether data sharing can improve long-term weight loss. This study will be the first to discern the mechanisms of action through which each type of data sharing may be beneficial, and elucidate conditions under which the benefits of data sharing may be maximized.

1. Introduction

Maintenance of healthy eating and activity behaviors is a challenge for adults with overweight or obesity who attempt behavioral weight loss (BWL) [1,2]. Participants in BWL programs commonly report that when meetings with coaches and fellow participants become less frequent or end, the reduction in available support and accountability makes continued adherence difficult [3,4]. Thus, one potential pathway for improving long-term outcomes is to create sustained sources of support and accountability.

Participants in BWL programs increasingly use digital tools to self-monitor body weight, physical activity (PA), and dietary intake [5–8]. As data collected with these tools can be shared with others, there is an opportunity for other parties, such as coaches, fellow participants, and/or friends or family members, to provide support and accountability for behavior change efforts that is informed by data collected from these devices [9–15]. Data sharing could enhance self-regulation, self-monitoring adherence, and motivation for behavior change by providing accountability for goals [16–20] and positive feedback that reinforces progress [21,22] and setting the stage for problem solving (Fig. 1). In fact, our team conducted a pilot study that demonstrated the promise of coach-based data sharing [23]. However, data sharing (particularly with fellow participants or friends/family members) is not yet a standard component of BWL programs and has not been evaluated as a means of improving long-term efficacy of BWL. The current study will address critical questions about data sharing and support in BWL by giving a coach, fellow BWL group members, and/or a friend/family member long-term access to participant digital self-monitoring data.

Fig. 1.

Conceptual model.

Coaches, fellow BWL participants, and friends/family members each have potential strengths and weaknesses as data sharing partners. Coaches may provide high quality feedback given their scientific expertise in behavior change [22]. On the other hand, coach surveillance could have an iatrogenic effect on long-term program engagement for some participants, e.g., avoiding coaching contact or self-monitoring because of a sense of shame. Data sharing with program peers might harness the power of shared goals [24,25] and adaptive social comparison to relevant peers [26–29]. In contrast, the quality of peer support could be variable and social comparison may be demoralizing for some participants. Friends or family members may have the advantage of a pre-existing relationship with frequent, in vivo contacts and mutual trustworthiness, benevolence, and respect, all of which are important components of effective supportive accountability [22,30] and a working alliance [31]. However, relationship quality and commitment to providing support may not be sufficiently high for all participants.

In the proposed study, adults with overweight/obesity will enroll in a remote, 24-month BWL program that includes group meetings, individual phone calls, and text messaging and use digital self-monitoring tools daily (wireless body weight scale, wearable PA sensor, and dietary intake app). A 2×2×2 factorial design will be used to test the effects of three types of data sharing partnerships: 1) Coach Share: The behavioral coach will regularly view digital self-monitoring data and address data observations. 2) Group Share: Participants will view each other’s self-monitoring data in their small-group text messages. 3) Friend/Family Share: A friend or family member will view the participant’s data via automated message.

Aim 1 is to test the hypotheses that Coach Share, Group Share, and Friend/Family Share will each improve long-term weight loss (primary outcome) and PA and calorie intake (secondary outcomes). Exploratory analyses also will determine if effects are partially or fully additive, antagonistic, or synergistic. Aim 2 is to test the following mediators of intervention effects: perceived supportive accountability, self-regulation, self-monitoring engagement, social comparison, and social support. Aim 3 is to determine how quality of the relationship with the coach, group members, and friend/family member may moderate treatment effects and predict data sharing efficacy. Qualitative data about participant experiences also will be collected to inform future research and dissemination efforts. As digital technology makes data sharing increasingly feasible, this study will have a major scientific impact by providing a clear indication of whether data sharing can improve long-term weight loss and treatment efficacy.

2. Methods

2.1. Participants

Adults (N = 320) with overweight or obesity will be recruited nationally, primarily by advertising on social media, and supplemented as needed by advertising on other digital media, radio stations, and publicizing the program in primary care practices. Recruitment will be conducted in four cohorts, with enrollment of the final cohort expected to be completed in first quarter of study year three, and data collection expected to be complete in the first quarter of study year five. Inclusion criteria include age 18–70 years; BMI 27–50 kg/m²; ability to engage in PA; and successful completion of all enrollment procedures. Participants must have a smartphone and internet access; financial assistance for this technology will be provided as needed. Participants also must identify one friend or family member age ≥ 18 who indicates willingness to serve in a support role. Exclusion criteria include medical or psychiatric conditions that may pose a risk to the participant, cause weight change, or limit ability to comply with the program; use of insulin or a medication that can cause significant change in weight; bariatric surgery history; currently pregnant or breastfeeding or planning to become pregnant in the next 24 months; weight loss ≥ 5% in the previous 3 months. In lieu of a run-in or wash-out period, participants will be required to attend a study visit to set up self-monitoring devices. At this visit, participants will review and sign the informed consent with a study staff member.

2.2. Study design

This clinical trial will utilize a 2×2×2 factorial design, with 8 different combinations of Coach Share, Group Share, and Friend/Family Share being “On or “Off” (see Table 1). Randomization to condition will be done by the study statistician at baseline, stratifying by BMI and age. Coaches and participants will not be blinded to treatment condition.

Table 1.

Factorial Design.

Condition	Coach Share	Group Share	Friend/Family Share
1	ON	ON	ON
2	ON	ON	OFF
3	ON	OFF	ON
4	ON	OFF	OFF
5	OFF	ON	ON
6	OFF	ON	OFF
7	OFF	OFF	ON
8	OFF	OFF	OFF

2.2.1. Trial registration

This study was pre-registered at clinicaltrials.gov (identifier: NCT05180448). The Drexel University Institutional Review Board approved all procedures.

2.3. Intervention

2.3.1. Intervention contact

All participants will receive 24 months of BWL, delivered remotely through a combination of group meetings (conducted via videoconferencing), coach phone calls, and in-app messaging. Table 2 details the schedule of intervention contact. Groups meet weekly for the first 3 months, followed by monthly calls or group sessions. In the 12-month pilot study for this study [23], clinically significant weight loss was observed with a similar cadence of contacts. By limiting weekly sessions to the first 3 months, and offering a combination of groups or coaching calls monthly thereafter, the cost and burden of the intervention may be limited while allowing for sufficiently frequent and sustained contact for the experimental intervention components to promote long-term weight loss. (One individual coaching call also will be provided to participants in the initial phase of the intervention, at week 6, in order to provide an early opportunity for coaches to address any concerns participants have about the intervention.)

Table 2.

Schedule of intervention contact.

Contact type	Month 1–3	Month 4–24
Group sessions	Weekly	1/quarter
Individual coach calls (15 min each)	Week 6**	2/quarter
Personalized text messages from coach	1/month	1/month
Small group messages*	1–2/week	1–2/week
Friend/family messages*	1/month	1/month

^*Note: Frequency of automated messages generated by the program are noted.

Automated messages are spaced throughout the month so that multiple automated messages are not sent on a single day. Participants may choose to send additional messages to their small groups or their friend/family members.

2.3.2. Group sessions

Each group will include approximately 14–16 participants. Within each group, approximately half of the participants will be assigned to Coach Share ON, half to Group Share ON, and half to Friend/Family Share ON. Thus, in each group, there will be approximately two participants assigned to each of the eight experimental conditions. This approach was selected to reduce the likelihood of group effects influencing outcomes. Coaches will give participants guidance about what information may and may not be shared in group in order to prevent contamination.

Sessions will begin with 20 min for coaches to conduct a brief, private, check-in with each participant in a breakout room, followed by a 70-min group meeting. There will be two coaches per group, one of whom is a senior coach with a masters or doctoral degree in psychology or a related field. The group meeting will begin with each participant reporting progress toward goals for behavior change, calorie intake, and PA. This will be followed by content designed to teach a new behavioral skill. The curriculum was adapted from the Look AHEAD [32] and the Diabetes Prevention Program protocols [33]. Core skills include self-monitoring, goal setting, stimulus control, and problem solving. Participants will set goals for calorie intake, based on weight, in accordance with standard balanced deficit diet guidelines. Nutritional guidance will allow for flexibility in how participants meet their calorie goals, while suggesting specific dietary changes that are likely to promote health, facilitate adherence to the calorie goal, and/or control hunger and cravings. Participants will be instructed to gradually increase their PA until they reach 250 min of moderate-to-vigorous PA per week. Participants will be encouraged to aim for 1–2 lbs. of weight loss per week until reaching a 10% weight loss. Throughout the program, coaching will focus on skills that facilitate long-term calorie restriction (i.e., a lower calorie intake compared to baseline) and high levels of physical activity, such that the content in later months of the program continues to be applicable to those with a goal of weight loss maintenance, or those who wish to lose more weight after reaching 10%, or those who have not yet reached a 10% weight loss.

2.3.3. Digital self-monitoring

Participants will be encouraged to engage in daily self-monitoring for the entirety of the program: weighing themselves with a smart scale, wearing a Fitbit wristband to track PA, and recording food intake in the Fitbit app. They will be told to sync their scales and PA sensors daily and encouraged to regularly view their data in the Fitbit app.

2.3.4. Experimental treatment components

As described next, ON and OFF conditions share most elements of treatment, such that data sharing is isolated as an experimental component as much as possible. Ten percent of coaching sessions (group and individual) will be recorded and coded for fidelity and contamination according to a structured rubric.

2.3.5. Coach communication

Coach Share OFF.

In Coach Share OFF, the behavioral coach has no access to device data. During individual check-ins and phone calls, coaches will ask participants about progress toward program goals. During group check-ins, coaches will hear participants’ reports of days of dietary and PA self-monitoring completed, average calorie intake, and minutes of PA. Goal setting and problem solving will be based on information from participants’ self-reports. Coaches will send personalized text messages monthly, referencing goals and/or strategies discussed in the previous coaching contact and provide encouragement for participant efforts in that area.

Coach Share ON.

If a participant is randomized to Coach Share ON, the coach will have access to self-monitoring device data via a customized, web-based dashboard. During private check-ins at the start of group meetings, individual coach calls, and monthly coaching text messages, the coach will share what they have observed from device data in terms of weight change, self-monitoring engagement, and diet and PA patterns, and will ask the participant to comment on their performance. The coach will provide praise and encouragement to reinforce adherence and relative improvements and will express concern about areas in which adherence is poor. Goal setting will be informed by device data, and the coach will set the expectation that progress will be monitored in an ongoing manner.

2.3.6. Group communication

Group Share OFF.

Within each BWL group, half of the participants will be randomized to Group Share OFF and will be combined into one small sub-group, with 5–8 participants, for text messaging in the app. Weekly, automated, prompts will encourage participants to share progress, challenges, or tips related to specific program goals and strategies (e.g., “In this program, we talk often about the importance of keeping tempting foods out of your home. How is that going for you?”). Participants will be encouraged to use group messaging as a source of social support, expanding beyond the suggested topics as they wish. Participants will be told that they should not report specific numbers (e. g., pounds of weight loss) from their self-monitoring in these messages. As is convention in BWL programs, participants will provide a self-report of their adherence to some program goals during group sessions.

Group Share ON.

The participants who are randomized to Group Share ON will also be combined into a sub-group, with 5–8 participants, for messaging in the app. Participants will receive the same automated prompts as are sent to participants in Group Share OFF. In addition, automated messages will summarize recent goal progress for each participant in the small group, reporting on the past month’s frequency for each type of self-monitoring, PA minutes, and percent weight change, as well as total weight change. The message also may highlight when a participant has a special accomplishment, such as a wearing their Fitbit for 100 days in a row. (See Fig. 2 for an example.)

Fig. 2.

Example of in-app automated messages.

2.3.7. Friend/family communication

Friend/Family Share OFF.

If a participant is randomized to Friend/Family Share OFF, a monthly, automated, message will be sent in the app to the index participant and his/her friend/family member. The message will provide education to the friend/family member about key principles of the program (e.g., explaining that PA is important for long-term weight control), and will prompt the index participant to let the friend/family member know how they can support them in their lifestyle modification efforts. The message content will be standardized and include no information about participant progress. Participants will be instructed to refrain from sharing any device data with friends or family members. Friend/family members will be invited to attend four webinars, each lasting 20–25 min (baseline, 6 months, 12 months and 18 months), that teach them how to provide support for the index participant’s lifestyle modification efforts.

Friend/Family Share ON.

If a participant is randomized to Friend/Family Share ON, all of the activities described in Share OFF occur, with two differences. The monthly, automated app message will, in addition to the standardized content, also report on the index participant’s past month’s frequency for each type of self-monitoring, PA minutes, and percent weight change, as well as total weight change. The webinars will integrate content teaching friends/family members to respond supportively to monthly data sharing reports.

2.4. Assessments

At months 0, 6, 12, and 24, participants will be instructed to complete a battery of self-report measures and daily self-monitoring of weight, PA, and food intake for 7 days. Self-report data will be housed on REDCap, a secure web-based data capture tool hosted at Drexel University [34,35]. Self-monitoring data collected from the Fitbit servers will be stored on an encrypted server that is only accessible by study staff. An abbreviated assessment also will be administered at 3 months, when participants will be asked to complete select measures and 7 days of weight self-monitoring. There will be no additional data collection after the completion of the 24-month assessment.

2.5. Primary outcome

2.5.1. Weight

Percentage of body weight lost at 24 months is the primary outcome variable. Participants will be instructed to use a Withings Body+ smart scale (stated accuracy to 0.1 kg [36]), to weigh themselves. To ensure weight accuracy, participants will be instructed to step on the scale in the morning before eating and after using the bathroom, wearing no shoes and minimal clothing, with the scale on a hard, flat, surface [37,38]. Seven days of measurement was chosen, rather than a single day, so that a mean could be calculated at each time point in order to minimize the influence of daily fluctuations of weight on this outcome.

2.6. Secondary outcomes

2.6.1. Physical activity

PA will be operationalized with two variables, both measured via Fitbit: 1) the sum of Very Active and Fairly Active minutes, and 2) daily step count. Participants will be instructed to wear the Fitbit for all waking hours for 7 days at each assessment point.

2.6.2. Calorie intake

At each assessment point, participants will be instructed to record all dietary intake in the Fitbit app for 7 days. The total calorie intake logged each day will be examined.

2.7. Mediators

The following constructs will be examined as mediators of intervention effects: perceived supportive accountability (Perceptions of Accountability scale [39] and Supportive Accountability Inventory-Adapted [40]), self-regulation (measure created by investigative team), social comparison (Iowa-Netherlands Comparison Orientation Measure-Modified [41]), and social support (Sallis Social Support for Diet and Exercise [42]). Self-monitoring engagement, another potential mediator, will be measured objectively as the number of days over the course of the program in which digital devices were used to measure weight, eating, and PA.

2.8. Moderators

Relationship quality will be examined as a moderator and measured with the Working Alliance Inventory - Short Revised [43], with separate versions assessing relationship with the coach, group members, and friend/family member, and with an additional measure of group and friend/family interactions created by the investigative team.

2.9. Qualitative data

Semi-structured interviews (n = 30) will be conducted at posttreatment, with purposive sampling to maximize diversity in participant characteristics and experiences with the weight loss intervention. Interviews will be recorded and transcribed verbatim. Two researchers will code each transcript. Themes and sub themes will be developed and reviewed by the research team.

2.10. Statistical analysis

Baseline characteristics will be compared for each ON versus OFF contrast, as well as for each of the 8 study arms. Key baseline variables that differ by condition will be considered for use as covariates.

2.10.1. Aim 1

This study will evaluate the hypotheses that long-term weight loss (primary outcome) and PA and calorie intake (secondary outcomes) will be superior in Coach Share ON vs. OFF; in Group Share ON vs. OFF; and in Friend/Family Share ON vs. OFF. We will model the pattern of change in weight over time using multilevel models [44,45]. The cross-level interaction between time and condition will be used to determine the effect of the partnership factor on the pattern of change in weight. Each of the three data sharing partnerships will be examined separately. Similar multilevel models will separately examine the effect of each of the three partnerships on PA and calorie intake. In addition to percent weight loss, two secondary, categorical weight outcomes will be compared across conditions: weight loss of ≥ 10% (Y/N), and weight loss of ≥ 5% (Y/N). We will conduct logistic regression to examine whether each of these outcomes differs by the partnership factor at each assessment point. As an exploratory aim, we also will test the interaction effects of the three data sharing partnerships by adding higher-order interaction terms to the afore-mentioned multilevel models.

2.10.2. Aim 2

Analyses will determine whether change in hypothesized mediators from baseline to 12 months explains the effect of condition on weight loss at 24 months, with the expectation that the Share ON participants will maintain a high level of the proposed mediator from baseline to 12 months, while Share OFF participants will show decreases in that variable. Coach Share, Group Share, and Friend/Family Share will each be examined separately. Using the mediation model outlined by Preacher and Hayes [46,47], we will examine the indirect effect of each hypothesized mediator using the bias-corrected bootstrap test [48–51].

2.10.3. Aim 3

To evaluate whether quality of the relationship with the coach, group members, or friend/family member moderated intervention effects, we will add it to the multilevel model described in Aim 1 and determine whether the condition effect on change in weight depends on this variable [51]. Relationship quality will be examined at baseline, and also will be examined separately at 6 and 12 months (controlling for weight loss up to that point). Statistically significant interactions will be interpreted by plotting simple regression lines for high and low values of relationship quality.

2.10.4. Qualitative data

Transcripts will be uploaded into NVivo qualitative software (QSR International Pty Ltd.) to support coding. Qualitative analyses will begin with a memoing process, where interviewers will write descriptive summaries and preliminary conclusions for each interview as they were conducted. Next, the team will develop code books. The majority of codes will be developed a priori based on the project goals and literature in this area, and additional emergent codes will be developed and added through discussion and consensus among the team members. Two researchers will code each transcript. Themes and subthemes will be developed and reviewed by other research team members to see if they were a true expression of participants’ experiences. Any interpretation or coding disagreement between the coders will be discussed with the project team until consensus is reached.

2.10.5. Attrition and power analyses

Imputation models that assume a trajectory of return over time to baseline weight, PA, and calorie intake levels will be used to handle missing data. We also will use Hedeker and Gibbons’ pattern mixture procedure [52,53] to examine whether dropouts influence the multilevel model results. If the missingness mechanism is related to the missing outcome itself, we will use sensitivity analyses to explore how robust our findings are with respect to a range of assumptions regarding missing data. Power calculations used the method described by Raudenbush [54–56] with the software Optimal Design. In our R21 study comparing LM with and without data sharing, the effect size was 0.55 for weight loss and 0.37 for PA. Based on these effect size estimates, we expect to detect an effect in a similar range (small to medium effect, d = 0.40) for weight loss and PA in the proposed study. Power analyses suggest that a sample size of 238 is required for 80% power with a significance level of 0.05 and four assessment points, assuming the ratio of the variability of level-1 coefficient to the variability of level-1 residual is at least one. This study is adequately powered with the proposed sample size of 320 with up to 25% attrition.

Tests of mediation will utilize bias-corrected bootstrapping, which demonstrates the best balance of statistical power and type I error [57]. Based on Fritz and MacKinnon’s [58] model for regression-based mediation models, under the assumption of a small to medium effect size for intervention with the mediator and a small to medium effect size for the mediator on outcome controlled for intervention, the required sample size is 148 to achieve 80% power in mediation models at a significance level of 0.05. [58] We will have 80% power to detect moderation if the standardized coefficient for the treatment by moderator interaction is at least 0.30 (accounts for at least 8% of the variance) in a model in which the total variance accounted for is at least 20%.

3. Discussion

The proposed study will use a 2×2×2 factorial design to determine whether long-term weight loss outcomes are improved when participants share their self-monitoring data (i.e., calorie tracking, PA, and weight change) with their coach (ON vs. OFF), fellow participants (ON vs. OFF), and/or a friend or family member (ON vs. OFF). The study design is innovative and methodologically rigorous. Although some past research has tested programs that involve data sharing with a coach, the current study is the first to isolate the effect of this variable to examine its causal influence in fully-powered, long-term trial. It is also the first of any study to test the efficacy of sharing BWL participants’ data with a designated friend/family member, as well as with program peers. The factorial design can examine the best combination of data sharing to maximize participant outcomes, allowing the possibility that adding partnerships may provide no added benefit or even be iatrogenic. By testing potential mediators of the experimental conditions, this study will improve our understanding of why data sharing partnerships may be beneficial. The moderation analysis will allow us to determine whether relationship quality influences the effectiveness of data sharing partnerships, which can be used to tailor future interventions that incorporate these strategies.

3.1. Limitations

Despite its many strengths, the study has several limitations. First, there is the possibility of contamination if participants in an OFF condition choose to share their data. For example, Friend/Family Share OFF participants could verbally report their weekly weight change to their designated support person, even though this type of spontaneous data sharing in OFF conditions will be discouraged. Second, the level of engagement and support provided by group peers and friend/family members cannot be experimentally controlled in the same way that it can be for coaches; for instance, some participants may not regularly read or send group text messages. Third, participants will not be blind to their treatment conditions. It is possible that participants assigned to one or more OFF conditions may have lower expectations of benefit from the program or feel disappointed about their condition assignment. Fourth, due to the intensive nature and burden of this intervention, participant engagement and adherence may decrease over time, limiting the opportunity for an experimental component to have an effect (e.g., if a participant stops attending sessions). Finally, we conducted a pilot study that provided an opportunity to identify and solve technical problems in data sharing and the study app. However, since the technological approach remains novel, there is a risk of temporary bugs which could diminish the impact of the intervention (e.g., if a coach is unable to view data before a group session).

4. Conclusions

Long-term weight loss maintenance is one of the greatest challenges in obesity treatment [59–62]. The proposed study will explore whether sharing performance data with others can provide the sustained support and motivation required to maintain long-term behavior change. The factorial design will inform which types of partnerships (and combinations of partnerships) show the greatest promise for further evaluation and dissemination. Data sharing is a low-cost and low-burden solution that, if effective, can be easily integrated into BWL interventions or be used as a simple, standalone intervention with potentially large effects.

Data availability

No data was used for the research described in the article.