Virtual fitness buddy ecosystem: a mixed reality precision health physical activity intervention for children

The study addresses the urgent need to promote and sustain healthy physical activity (PA) behaviors during young to middle childhood (ages 6–11), a critical period for establishing lifelong habits that protect against obesity and related chronic disease. Prior research shows most youth do not meet PA guidelines and experience steep declines in PA with age, partly due to increased sedentary time in school settings. Technology-mediated interventions (e.g., exergames) show promise but often produce modest, short-lived effects. Guided by self-determination theory (SDT), which emphasizes autonomy, competence, and relatedness as drivers of intrinsic motivation, the authors posit that interventions should enable children to set self-determined goals, experience mastery through tailored feedback, and receive social support to internalize PA behaviors. The precision health approach leverages digital technologies to personalize intervention content, provide timely feedback, and scale social connections. The research question tests whether a mixed reality, precision health intervention—the Virtual Fitness Buddy (VFB) ecosystem—can increase PA and reduce sedentary behavior among children in afterschool settings compared with an active control. The VFB integrates a virtual dog agent, wearable sensors, and a mixed reality kiosk to facilitate goal setting, tailored reinforcement, and parent support.

The paper situates the work within literature on technology-mediated PA interventions for children, particularly exergames, which can initially boost PA but often fail to sustain it. SDT literature suggests extrinsic rewards (e.g., points, badges) can undermine sustained motivation once rewards cease, highlighting the need to fulfill autonomy, competence, and relatedness needs. Precision health frameworks advocate tailoring interventions via wearables, sensors, and virtual agents to enhance intrinsic motivation and social connectedness cost-effectively. The study also references evidence on the importance of light-intensity PA and 24-hour movement guidelines, suggesting that displacing sedentary time with light PA can yield meaningful health benefits in youth and may be more achievable within unstructured play contexts than focusing exclusively on MVPA.

Design: Multisite, cluster-randomized, controlled trial across 19 YMCA-affiliated afterschool sites in metropolitan Atlanta over two academic-year cohorts (each 6 months). Randomization was at the afterschool program level in matched pairs (matched by percentage of children receiving free lunches). The trial was non-masked to participants, staff, and researchers. Registered at ClinicalTrials.gov (NCT03524183); IRB approved by the University of Georgia. Participants: Children in grades 1–5 (ages 6–11) enrolled in afterschool programs, able to perform moderate-intensity PA without assistance, and attending most days. Parent–child dyads provided consent/assent. Recruitment yielded 257 children in cohort 1 and 165 in cohort 2 (total randomized N=422). For accelerometer-based ITT analyses, 303 children had valid data (≥8 hours wear on at least one day at each of the 3 time intervals). Intervention (VFB ecosystem): Children interacted with a custom virtual dog via a motion-tracking mixed reality kiosk. Fitbit wearables synced step/PA data to the kiosk; the virtual dog provided tailored, precise feedback. As children met self-determined weekly PA goals, the dog’s appearance/behavior signaled vicarious reinforcement (slimmer, faster, more energetic), and children could play with the dog for limited periods (~10 minutes) via gestures and movement. The kiosk automatically notified parents of progress; parents could send encouraging messages delivered through the kiosk. The design aimed to foster autonomy (self-set goals), competence (tailored feedback and mastery), and relatedness (support from dog and parents). Kiosk content discouraged prolonged on-site PA, encouraging real-world activity. Control: Children used a computer system to set PA goals without support/feedback from the virtual dog or parents (active control). Procedures and timeline: Baseline (Time 1), 3-month (Time 2), and 6-month (Time 3) assessments included surveys and body composition (height, weight, waist). All children received Fitbits for ongoing ecosystem engagement (intervention) or goal-setting (control). To validate Fitbit measures (reporting elsewhere), up to 20 children per site wore ActiGraph GT9X accelerometers for 7 continuous days (waking hours; removed for water activities) at each time point; parents logged wear times and non-wear reasons. Kiosks were retrieved at the end of Time 3. System usability enhancements between cohorts 1 and 2 improved syncing and login (no change to virtual dog behavior) to increase fidelity and uptake. Outcomes and measurement: Primary outcomes were daily minutes in sedentary, light, moderate, vigorous PA, and daily steps, derived from ActiGraph data (Firmware v1.7.1; ActiLife v6.13.4; 10-second epochs). Intensity classification used Evenson cut-points; wear time estimated using the Choi algorithm (non-wear: ≥90 minutes of zero counts). Valid days required ≥8 hours wear. Playtime was defined as daily duration of child–kiosk interaction with the virtual dog. Due to COVID-19 closures after Time 3 of Cohort 2, analyses cover the first 6 months for both cohorts. No adverse events were actively monitored; none were reported. Statistical analysis: Intent-to-treat (ITT) analyses pooled cohorts; sensitivity analyses stratified by cohort, baseline MVPA (<45 vs ≥45 minutes/day), and dog ownership. Hierarchical linear mixed models estimated treatment-by-period interactions (baseline, 3 months, 6 months) for each outcome, adjusting for hours of wear, child overweight (BMI%ile ≥85), sex, race/ethnicity, and age; and parent overweight, sex, and education. Random intercepts for school and child (nested within school) and a random slope for observation day within child accounted for clustering and longitudinal correlation. Cluster-robust standard errors were used; two-sided alpha=0.05. Intraclass correlation coefficients (ICCs) were reported for child- and school-level clustering. Descriptive kiosk engagement metrics summarized adherence and exposure by site and cohort.

• Sample and balance: 303 children with valid accelerometer data; treatment and control arms were balanced on demographics and anthropometrics. Baseline MVPA near guidelines (Treatment: 53.9 min; Control: 57.7 min).
• MVPA (primary): No significant overall treatment-by-time interaction (p=0.092). Treatment effects: Time 1 −3.8 min (95% CI −8.5, 0.9), Time 2 +3.9 min (95% CI −2.5, 10.2), Time 3 −6.0 min (95% CI −14.5, 2.4). Within-arm, MVPA rose in Treatment by ~3.5 min from Time 1 to Time 2 and declined by ~4.2 min in Control over the same interval.
• Vigorous PA: Interaction p=0.150; treatment effects not significant across periods.
• Moderate PA: Interaction p=0.071; no significant differences across periods.
• Light-intensity PA: Significant treatment-by-time interaction (p=0.001). Treatment effects: Time 1 −2.6 min (95% CI −16.7, 11.5), Time 2 +17.4 min (95% CI 7.7, 27.2), Time 3 +2.5 min (95% CI −16.4, 21.5).
• Sedentary time: Significant treatment-by-time interaction (p=0.005). Treatment effects: Time 1 +6.2 min (95% CI −10.5, 22.9), Time 2 −21.7 min (95% CI −34.1, −9.2; p<0.001), Time 3 +3.4 min (95% CI −22.1, 29.0). Reductions at Time 2 primarily converted to light-intensity PA.
• Daily steps: No significant interaction (p=0.17).
• Subgroup: Baseline MVPA <45 min/day (n=99 total; n=50 Treatment) showed stronger sedentary reduction at Time 2 (−37.3 min/day) than those with higher baseline MVPA (−23.9 min/day). Among lower MVPA children, 80% of reduced sedentary time converted to light PA (vs 66% in higher MVPA).
• Dog ownership: About one-third owned a dog (n=106). At Time 2, light PA effects were larger for dog owners (+26.6 min) than non-owners (+11.5 min). At Time 3, effects sustained/increased for dog owners (+30.5 min) but not for non-owners (−11.5 min).
• Implementation fidelity: Cohort 2 showed improved effects versus Cohort 1. From Time 1 to Time 2, MVPA gain in Treatment vs Control was +7.6 min/day (95% CI 3.1, 12.0) in Cohort 2 vs +2.1 min/day (95% CI −8.3, 12.6) in Cohort 1. Cohort 2 achieved MVPA above 60 min/day at Time 2 (61.8 min; 95% CI 60.1, 63.6). Kiosk availability and engagement were higher in Cohort 2 (available 34.8% of days vs 24.0% in Cohort 1; >100 s playtime on >70% of active days).

The VFB ecosystem, a precision health intervention grounded in self-determination theory, effectively reduced sedentary time and increased light-intensity PA over the first three months compared with an active control. Although MVPA gains were not statistically significant overall, the intervention appeared to counteract the typical age-related decline in PA by sustaining baseline levels through six months, rather than allowing the expected deterioration seen in this age group. Conversions of sedentary time predominantly to light-intensity PA align with contemporary 24-hour movement guidelines emphasizing all intensities. Subgroup analyses suggest greater benefits for children with lower baseline MVPA and those with pet dogs, indicating potential moderators that can guide targeted implementation. Cohort 2 results underscore the importance of implementation fidelity; usability enhancements and greater kiosk availability were associated with stronger treatment effects, including achieving guideline-level MVPA at three months. Collectively, these findings support the relevance of integrating wearable sensors, mixed reality agents, and parent-mediated support within afterschool settings to facilitate autonomy, competence, and relatedness for sustained behavior change. The approach offers a scalable, cost-efficient model for promoting unstructured, enjoyable PA and reducing sedentary time in children.

This cluster-randomized trial demonstrates that a mixed reality, precision health intervention integrating a virtual pet, wearables, and parent feedback can meaningfully reduce sedentary time and increase light-intensity PA over three months among children aged 6–11, with maintenance of baseline PA thereafter. Effects were stronger with improved system fidelity, among children with lower baseline MVPA, and among dog owners. The VFB ecosystem offers a scalable, cost- and labor-efficient solution that can be embedded in existing afterschool infrastructures for broad public health impact. Future work should focus on sustaining gains beyond three months by increasing content variety and challenge, expanding access beyond afterschool calendars, extending observation across seasons and academic years to disentangle seasonality, and further automating components to reduce staff burden and enhance adherence.

• Short-term attenuation: Gains at three months were not sustained at six months; MVPA treatment effects were not statistically significant overall.
• Seasonality: Months 3–6 overlapped with winter, which typically reduces PA and increases sedentary behavior, potentially confounding effects.
• Access constraints: Kiosk access depended on afterschool schedules; closures (e.g., holidays) limited interactions. Kiosk availability and engagement varied across sites, especially in Cohort 1.
• Non-masking: Participants and staff were aware of intervention condition, introducing potential performance biases.
• COVID-19 impact: Data collection limited to first six months; afterschool programs shut down after Time 3 of Cohort 2.
• Generalizability: Conducted within YMCA programs in one metropolitan area; may limit external validity.
• Measurement scope: ActiGraph subsample design (up to 20 per site per time point) and requirement for ≥8 h valid wear may influence representativeness; no active adverse event monitoring (none reported).