Overview
As the global population ages, innovative approaches are needed to promote healthy aging. This randomized controlled trial evaluates a smartphone-based platform that remotely assesses motor fitness and delivers AI-generated, personalized exercise programs for older adults. The goal is to improve balance, flexibility, and strength without requiring laboratory facilities or professional supervision.
Rationale and Innovation
Traditional exercise guidelines often adopt a one-size-fits-all model, with limited guidance on nuanced components such as balance, flexibility, and targeted strength. This study leverages a simple smartphone equipped with accelerometer and gyroscope sensors to remotely measure key motor fitness components across multiple movement tests. A machine learning system then crafts individualized exercise programs delivered directly as video content on the phone, enabling high-fidelity personalization in a home setting.
Methods at a Glance
In a three-arm randomized design, 239 healthy community-dwelling older adults (aged 65+) in Israel were assigned to one of three groups for 8 weeks: (1) a personalized exercise program, (2) an active-control following WHO guidelines, and (3) a no-intervention control. Assessments were conducted at baseline, 4 weeks, 8 weeks, and a 12-week follow-up. Primary outcomes focused on motor fitness: balance (static and dynamic), flexibility (upper body), and strength (upper body and sit-to-stand).
How the Platform Works
Participants first complete a remote baseline assessment via the app. The tests are conducted using smartphone sensors attached with a simple strap, capturing five digital markers per balance test and specific measures for flexibility and strength. The data feed into a machine learning model that constructs a unique fitness profile for each participant and selects an appropriate set of exercises across three categories: balance and lower body, upper body flexibility, and upper body strength. Two difficulty levels—A (easy) and B (advanced)—are matched to each component.
Intervention Details
Personalized group: Exercise 5 days/week for 8 weeks, with videos delivered to the smartphone and tailored to the individual’s profile and progress. The program emphasized balance, upper body flexibility, and strength, with adjustments based on ongoing DM (digital marker) feedback.
Active-control group: Participants followed WHO recommendations (150–300 minutes of aerobic activity weekly plus 3 days of multicomponent training focusing on balance and strength).
Control group: Usual activities with no structured program during the study; access to a personalized plan was offered after study completion.
Key Findings
The personalized exercise group demonstrated superior improvements in dynamic balance, arm flexibility, and upper body strength compared with both control conditions. Notably, improvements were evident within 4 weeks in several measures, and benefits persisted at the 8-week mark. Between-group differences favored the personalized group, particularly for dynamic tandem balance and arm-related flexibility and strength metrics. Some balance measures did not show significant differential gains, underscoring the multifactorial nature of balance control.
Adherence emerged as a critical factor. Participants achieving at least 1.5 sessions per week showed meaningful gains, illustrating that even modest engagement with a tailored, remotely delivered program can yield tangible improvements. However, detraining effects were observed after discontinuation of activity, highlighting the need for ongoing exercise to maintain gains.
Clinical and Practical Implications
This study supports the feasibility of distributing personalized, AI-guided exercise prescriptions through a widely accessible device. By removing barriers such as laboratory assessments and in-person supervision, the platform can scale to broader populations, including those with mobility limitations or in remote areas. The approach also supports continuous optimization of exercise prescriptions as more data accumulate, enhancing precision medicine for aging populations.
Limitations and Future Directions
Limitations include sample bias toward higher socioeconomic status and the restriction to two difficulty levels per movement component. Future work should expand movement profiles, include additional fitness domains, and test the platform across diverse populations, including those with mobility or cognitive impairments. Integrating genetic, clinical, and behavioral data could further refine personalization, while larger-scale deployments would assess real-world effectiveness and cost-effectiveness.
Conclusion
The trial demonstrates that a smartphone-based platform enabling remote motor fitness assessment and AI-generated personalized exercise programming can outperform generic WHO-guideline approaches and no intervention in improving balance, flexibility, and strength among older adults. The approach holds promise for scalable, home-based health promotion in aging populations.
