A total of 1 in 3 children in the United States is either overweight or obese [1,2]. Obese children and adolescents are at risk for carrying their excess weight into adulthood and developing multiple comorbidities, including diabetes and coronary disease as adults [3,4]. Clinical management guidelines and treatment algorithms call for a staged approach to the treatment of overweight and obese children, which aims to promote healthy lifestyle changes through behavioral counseling [4,5]. Effective behavioral interventions for pediatric obesity involve multiple components focused on promoting healthy eating and exercise habits [6]. The most effective interventions involve supporting both children and parents to set goals, incorporate stimulus control, utilize problem solving, and participate in self-monitoring while working to achieve behavior changes [6]. The US Preventive Services Task Force (USPSTF) recommends primary providers to either provide or refer children with obesity to comprehensive intensive behavioral interventions aimed at decreasing excess weight and improving overall weight status [6]. However, child participation in comprehensive intensive behavioral interventions and clinical weight management programs is often low, with considerably high program attrition [7-9]. Common barriers to participation in weight management programs identified by families include concerns regarding affordability, inflexible scheduling, conflicts with other activities, time commitment, distance and transportation, and misalignment between expectations and program services [8,10,11]. Conversely, facilitators of participation for children and families may include tailored treatment plans and individualized health coaching [10-12]. Mobile health (mHealth) and telehealth technologies may provide a unique opportunity to overcome barriers to participation in obesity treatment by providing individualized interventions on a family’s timeline and in their home environments [13-15].

Among obesity treatment trials for adults, mHealth tools appear to successfully assist patients in managing comorbidities, such as diabetes, improve physical activity and dietary behaviors, and achieve meaningful weight loss [16-22]. mHealth interventions in children and adolescents have been found to be effective at improving health behaviors and health outcomes across a wide reach of conditions [23]. In obesity treatment trials for children, the effect of mHealth tools is less clear, as they have been primarily studied as a component of larger, multifaceted interventions [13,15,24,25]. mHealth tools incorporated into obesity prevention and treatment trials for children vary in their ability to improve weight outcomes [15,24,25]. However, mobile technology has been shown to be well accepted, feasible, and effective at supporting self-monitoring and promoting changes of physical activity and dietary behaviors [13,24-26]. However, overall, there is limited evidence for the efficacy of mHealth interventions as stand-alone treatment modalities for pediatric obesity and weight management [13]. Although there are a number of commercially available mobile apps targeting weight-related physical activity and dietary behaviors in children, reviews of commercial apps have found that most lack high-quality information, include only a few behavior change techniques (BCTs), and are not rooted in evidence-based behavior change theories [27,28]. As a result, some have suggested the need to rigorously evaluate commercial and stand-alone mHealth interventions aimed at promoting health behavior change among overweight or obese children [13,27,28].

In this study, we aimed to assess the engagement of overweight or obese children with a commercially available mHealth tool (Kurbo), which provides individualized health coaching and self-monitoring support designed to improve diet and physical activity behaviors. The primary aim was to describe and compare the engagement of participants with health coaching sessions, as a condition of their commitment period. The secondary aim was to examine the association between coaching sessions received and the change in a participant’s weight status over time. We hypothesized that participants with longer commitment periods would engage in both more coaching sessions and have a trend toward greater weight loss.

This was a retrospective cohort study of participants in a commercial, mobile app–based platform and program (Kurbo) designed to promote health behavior change and weight management through self-monitoring and health coaching support. The research team and investigators had no role in the development of the mobile app platform or the creation and delivery of program content.

The Kurbo mobile app platform (Figure 1) and program was designed to promote behavior change and encourage healthy lifestyle choices [29]. The program content and health coaching incorporate multiple BCTs, consistent with established taxonomy for behavior change interventions [30]. The BCTs emphasized in the program are linked to multiple theoretical frameworks, including the theory of reasoned action, theory of planned behavior, social cognitive theory, and control theory, as well as operant conditioning and the information motivation behavioral skills model [31].The program design was also informed by a model of supportive accountability, which emphasizes the essential role of human support in mHealth interventions [32]. The mobile app and program include 2 primary components: (1) self-monitoring of eating and physical activity behaviors through a mobile app interface and (2) individualized coaching sessions through video chat.

The self-monitoring component of the program employs the BCTs of self-monitoring of behavior and monitoring of outcomes of behavior [30]. Participants are encouraged to use the mobile app to log their daily food intake, using food categories adapted from the evidence-based traffic light system [33,34]. The traffic light system categorizes foods into 3 groups: unrestricted healthy green-light foods, less healthy yellow-light foods that should be eaten with caution, and unhealthy red-light foods that should be avoided [34]. The aim of this approach to eating behaviors is to encourage participants to gradually increase consumption of healthy foods (green lights) and decrease unhealthy foods (red lights) over time. This approach incorporates behavior substitution and habit formation [30]. Participants are also asked to self-monitor their physical activity behavior by logging the duration of activities in the mobile app, while working toward a goal of 60 min of moderate-to-vigorous physical activity each day.

The individualized coaching sessions component is provided by individuals who are hired and trained as coaches by Kurbo. Participants are paired with the same coach for the duration of their participation with the program, which aims to provide social support and accountability [30,32]. Coaches monitor the participant’s dietary and physical activity behaviors through a Web-based dashboard, which serves to reinforce self-monitoring and allows for feedback of behavior [30]. Each coaching session lasts about 15 min, and it is made available on a weekly basis. Coaching sessions emphasize review of past behavior and outcome goals, as well as support future outcome and behavior goal setting [30]. Coaches encourage self-talk and identification of behavioral cues, as well as assist with tailored problem solving and action planning [30]. For example, coaches may discuss environmental cues and identification of triggers to eat red-light foods while supporting goal setting and action planning for choosing more green-light foods. Additional topics addressed during coaching sessions may include understanding food labels and portion sizes. After each coaching session, participants receive an email from their coach, with praise for goals met and a tailored plan regarding goals set for the next week. Of note, parents are strongly encouraged to participate in coaching sessions; however, the program is not prescriptive regarding parental involvement. Similarly, the program allows for individually tailored coaching, but it is not specifically designed to address family dynamics or the age of participant.

In addition to coaching sessions, participants are also able to contact their coach between coaching sessions, via short message service text messages, email, or in-app messaging. Independent of the Kurbo mobile app platform, participants also have access to supplementary resources, including an emailed e-workbook, biweekly email newsletter, physical activity demonstration videos, blog posts, and downloadable healthy-eating cookbooks. These supplementary resources highlight BCTs by providing instructions on how to perform behaviors, support restructuring of the physical environment, and allow for social comparison. Finally, the program also makes use of the BCT of a behavioral contract [30].

The study examined a retrospectively identified cohort of participants who initially utilized the Kurbo mobile app and program between March 15, 2015 and March 15, 2017. Participants were not recruited for the purpose of conducting the study. Deidentified participant data from the Kurbo data registry were provided to the investigators for the purpose of the study. The study received an exemption from the Stanford University School of Medicine Institutional Review Board. All data were independently reviewed and inspected by the research team to confirm that the predetermined inclusion and exclusion criteria were met before analysis.

Study participants were voluntarily subscribed or assigned to 1 of 3 commitment periods: 4 weeks, 12 to 16 weeks, or 24 weeks. Each participant was supported by either an employer-benefited plan, a health insurer–benefited plan, or a self-paid plan. The cost of the program is covered by either the parent (self-pay), a parent’s employer, or a family health insurance plan. Self-pay rates are dependent on the commitment period. Participants in self-paid plans voluntarily chose from 1 of 3 commitment periods: 4 weeks, 12 weeks, or 24 weeks. Those in employer- or health insurer–benefited plans were contractually assigned to 1 of 2 commitment periods: 12 weeks or 16 weeks. Of note, there were only 16 participants in the 16-week commitment period, and for analytic purposes, these participants were combined with the 12-week period to form a 12- to 16-week commitment period group. All participants had the ability to renew or change their plan at the end of the initial commitment period; however, data regarding renewals or changes were not available for analysis.

Descriptive statistics were used to compare baseline characteristics (age, age group, sex, BMI percentile, %BMIp95, weight category, obesity class, and payment source), primary outcome (participant engagement), and secondary outcome (change in weight status) across the 3 commitment periods. Categorical variables were expressed as absolute values and corresponding percentages. Normally distributed continuous variables are reported as a mean with standard deviation (SD). The continuous engagement measures analyzed had nonnormal distribution patterns, each of these are reported as a median with interquartile range (IQR). Differences of measures across commitment period groups were explored using Chi-square tests for categorical variables and analysis of variance for normally distributed continuous measures. Similarly, differences across commitment periods for nonparametric continuous measures were analyzed using Kruskal-Wallis tests. Significance of change in weight status within commitment periods was analyzed using paired two-tailed t tests.

Multivariable linear regression models were constructed to examine 2 sets of associations: (1) between the primary outcome (number of coaching sessions) and each commitment period (reference of 24-week period) and (2) between the primary outcome (number of coaching sessions) and the secondary outcome (change in %BMIp95). Each multivariable model included adjustment for significant baseline differences in age group and sex. A sensitivity analysis, excluding involuntary participants (ie, health plan or employer supported), was performed to isolate differences associated with voluntariness of commitment period (Multimedia Appendix 1). All analyses were conducted using SAS Institute Inc software (SAS University Edition/SAS Studio 3.71).

Of the 3242 participants assessed for eligibility, 1579 met the inclusion criteria. Of those, 305 participants were excluded for being outside the age range, normal weight at baseline (79), missing baseline data (46), or data measurement error. This yielded a final analytic sample of 1120 study participants, displayed in (Figure 1).

The baseline characteristics for the study sample by commitment period are displayed in (Table 1). Overall, 292 participants were in the 4-week commitment period, 690 participants were in the 12- to 16-week group, and 138 participants were in the 24-week commitment period. Mean age at baseline was 12 years (SD 2.5), and most participants 68.04% (762/1120) were female. There were no statistically significant differences across the 3 groups in age or sex. The majority of participants 76.61% (858/1120) were categorized as obese, with mean BMI percentile (SD) of 96.6 (3.1) and mean (SD) %BMIp95 of 114.5 (16.5). Children in the 24-week group were more likely to be classified as obese, when compared with those in the 4- and 12-week groups (118/138, 85.5% vs 218/292, 74.7% and 522/690, 75.7%, respectively, P=.03). The predominant payment source (743/1120, 66.34%) was self-pay. The distribution of baseline characteristics for the study sample by age group are displayed in Multimedia Appendix 1. Except for payment source, all baseline characteristics differed significantly across age groups.

The engagement of participants with the mobile app program, compared across commitment periods, is displayed in (Table 2).

The primary outcome of median number of coaching sessions received was 8 (IQR 3-15) for the 4-week group, 9 (IQR 5-12) for the 12- to 16-week group, and 19 (IQR 11-25) for the 24-week group (P<.001). Overall, the median number of coaching sessions per participant was 9, with an IQR of 5 to 15. The median (IQR) values for other engagement measures were as follows: participation period, 15 weeks (IQR 12-30); the number of coaching messages, 3 (IQR 0-10); the number of logged dietary events, 174 (IQR 83-325); and the number of logged physical activity events, 42 (IQR 15-91). Median weeks of participation differed across commitment periods: 16 weeks (IQR 8-36) for the 4-week group, 14 weeks (IQR 12-22) for the 12- to 16-week group, and 30 weeks (IQR 22-51) for the 24-week group. Overall, program retention was high, with 79.91% (895/1120) of participants remaining engaged with the program for at least the duration of their commitment period. Program retention across commitment periods was 92.5% (270/292) for the 4-week group, 76.8% (530/690) for the 12- to 16-week group, and 68.8% (95/138) for the 24-week group (P<.001). In addition, the engagement of all participants stratified by age group is shown in (Table 3). There were no statistically significant differences in engagement by age group.

Results of unadjusted and adjusted models for the primary outcome (number of coaching sessions per participant) are displayed in Table 4. After adjustment for child sex and age group (with the 24-week group as reference), the 12- to 16-week group was associated with the fewest number of coaching sessions per participant, with a beta-coefficient of –9.34 (95% CI –11.30 to –7.39), in contrast to the 4-week group, with a beta-coefficient of –8.03 (95% CI –10.19 to –5.87). The results of sensitivity analyses restricted to only self-pay (voluntary) participants are included in Multimedia Appendix 1. In this subpopulation, the 4-week group was associated with the fewest number of coaching sessions per participant, with a beta-coefficient of –8.06 (95% CI –10.56 to –5.56), in contrast to the 12-to 16-week group, with a beta-coefficient of –6.02 (95% CI –8.48 to –3.56).

Within each commitment period, the mean change between baseline and endpoint for %BMIp95 was –5.4 (95% CI –6.2 to –4.5) for 4 weeks (P<.001), –4.8 (95% CI –5.3 to –4.3) for 12 to 16 weeks (P<.001), and –6.9 (95% CI –8.3 to –5.6) for 24 weeks (P<.001). Compared across age groups, the mean change of %BMIp95 was –5.6 (SD 7.9) for 5 to 11 year olds, –4.7 (SD 5.9) for 12 to 14 year olds, and –5.2 (SD 5.6) for 15 to 18 year olds (P=.09). Adjusting for age group and sex within each commitment period, the beta-coefficient per coaching session was –0.25 (95% CI –0.32 to –0.18) for the 4-week group, –0.16 (95% CI –0.21 to –0.11) for the 12-week group, and –0.26 (95% CI –0.34 to –0.18) for the 24-week group (P<.001). The multivariable model of all participants, adjusted for initial commitment period, age group, and sex, demonstrated an overall beta-coefficient decrease of –0.21 (95% CI –0.25 to –0.17) in %BMIp95 per each coaching session received (P<.001).

This retrospective study described the engagement of a large cohort of children and adolescents, with a multicomponent mobile app–based comprehensive behavioral program aimed at promoting healthy dietary and exercise lifestyle behaviors. Unlike traditional behavioral interventions and clinical weight management programs, which largely rely on in-person visits and sessions [6,40,41], mHealth programs, such as the one studied, enable participants to self-monitor health behaviors and receive health coaching at their own pace. As such, the findings of this study generally add to the growing evidence base for mHealth tools, more specifically for mobile app–based comprehensive behavioral programs to support health behavior change for overweight or obese children and adolescents.

Our findings of overall engagement with a median of 9 (IQR 5-15) coaching sessions during the participation period is notable for an mHealth program. This level of engagement, although considered low intensity by USPSTF criteria, is comparable with contact levels of in-person weight management programs [40,41]. We also documented consistent levels of participant engagement with other measures, such as self-monitoring of physical activity and dietary habits. The documented engagement of participants with both individualized coaching and self-monitoring is an important finding, given that these components represent behavioral change techniques that are known to be effective in managing pediatric obesity [42]. Furthermore, overall program retention with this mHealth program was high, and attrition was 20.9%, which is considerably lower than attrition rates of between 37% and 41%, reported for traditional in-person (non-mHealth) weight management programs [8].

Finally, in this observational study, we found a significant association between the number of coaching sessions and the change in self-reported weight status during participation in the program. This association suggests that greater exposure to coaching sessions in this study was correlated with increased weight reduction. USPSTF analyses of intensive in-person interventions suggest a dose-response relationship between intervention hours received and beneficial changes in weight, with effective programs requiring at least 26 contact hours [6]. Although this study design was unable to capture a participant’s absolute contact hours with the app or coaching sessions, the observed association between weight change and coaching sessions suggests a possible lower dose or lower threshold effect for this mHealth intervention, when compared with traditional in-person interventions. However, further research is necessary to validate the nature and magnitude of this association, as well as to rule out other explanations, including selection bias or reporter bias.

The study has limitations common to other studies of digital health and mHealth interventions [15]. As an observational retrospective study design, the study also lacked an independent control or comparison group. The study is subject to some reporting bias, as all measures, including anthropometrics, were self-reported by participants. The study is also subject to selection bias, both because enrollment and participation were self-directed and because full functionality required access to an internet-connected mobile device. Another limitation was unmeasured confounders, including race/ethnicity, parental educational attainment, and household income, which limits adjustment and generalization, particularly for low-income populations [2]. However, the inclusion of employer-benefited and health plan–benefited participants, which may have included both commercial and government sponsored plans, likely increased the heterogeneity of the sample. Finally, the study as designed was unable to fully account for other factors that may influence intervention fidelity, including quality of coaching sessions, role of parents, clustering by coach, or use of other resources. Specifically, the study could not account for the degree of parental involvement and supervision among younger participants as compared with older participants. Nonetheless, the findings do not support a significant difference between age groups.

Still, rigorous and independent studies of digital health and mHealth interventions are limited, and this study meaningfully contributes to the literature on digital health and mHealth interventions focused on obesity treatment and prevention. First, the intervention was comprehensively designed with multiple components that have been shown to be effective strategies for weight management and behavior change in pediatric populations [6,30,33,34,42]. This is in contrast to the often limited BCTs employed in other mHealth tools [27,28]. Second, the intervention leveraged effective BCTs to support self-guided behavior change and utilized health coaches in a family context [24]. Third, the study population included a broad distribution of age groups, including preadolescent children, adolescents, and late adolescents. In fact, the age, sex, and weight characteristics of the study population are similar to that of traditional in-person medical weight management programs [40,41]. In addition to obese participants, the intervention also included overweight participants, which suggests generalizability of the findings beyond this study. Finally, it is notable that engagement with the program and change in weight status was consistent across age groups that include children, preadolescents, and adolescents.

Our findings have implications for clinical care, population health, and public policy. Clinicians providing obesity treatment may consider the incorporation of mHealth programs, such as the one studied here, as an adjunct to clinic visits and traditional medical management strategies. Health care systems aiming to improve population health management efforts might find these types of mHealth solutions more accessible for providing access to care for patients in rural areas where availability of providers may be limited or to patients in urban areas, who may be restricted by long commute times or have limited transportation options. Finally, public health leaders and policy makers may be encouraged by the role that emerging digital technologies could play in addressing obesity at the community level.

This study of a mobile app–based health behavior change and health coaching program among a large cohort of overweight and obese participants demonstrated high participant engagement. Increased engagement with coaching sessions was associated with longer voluntary commitment periods. Overall program retention was higher than that reported for similar in-person intensive behavioral interventions and weight management programs. Participant engagement with coaching sessions was associated with decreases in weight status (%BMIp95). Taken together, these findings highlight the potential of mHealth platforms as a promising model for delivering behavioral interventions that support weight management and behavior change for overweight or obese children and adolescents.