Clinicians define design rules for children’s upper-limb rehabilitation games

An iterative, clinician-led study has identified a set of practical design principles for turning routine upper-limb rehabilitation exercises into video games for children. The project developed two prototypes around equipment already used at a rehabilitation center in Morelia, Mexico: Hungry Hamster, which responds to movements on a finger ladder, and Treasure Hunter, which responds to rotations of a shoulder wheel. Three rounds of focus groups with physicians, a physical therapist and technical staff shaped the systems before 14 children tested them. Children generally preferred the gamified versions of their exercises, but the small, exploratory study did not test whether the games improve arm function or long-term rehabilitation outcomes.

The researchers treated the work as a user-centered development study rather than a clinical trial. The underlying devices—a finger ladder and shoulder wheel—are general-purpose tools for practicing upper-limb motor control, range of motion and movement coordination. Their use for each child was determined by the treating therapist, according to prescribed rehabilitation goals and functional abilities. Clinicians had identified the conventional exercises as repetitive and potentially insufficiently engaging, so the developers instrumented the existing equipment instead of asking children to substitute an unrelated commercial game for therapeutic movement. This approach was intended to preserve a direct connection between what a child physically does and what happens on the tablet screen.

Hungry Hamster used capacitive sensors attached to an ESP32 microcontroller to detect touches on eight of a finger ladder’s 32 rungs. Each sensor was calibrated to distinguish contact from non-contact, and the controller registered progress only when instrumented rungs were activated in ascending order. If a child skipped a rung, the virtual hamster stopped rather than advancing. The controller transmitted the identity and detection time of each rung to the tablet, allowing the software to calculate the interval between adjacent rungs, total repetition time and maximum height reached. Treasure Hunter used a different input method: an MPU6050 inertial measurement unit on the shoulder wheel sent angular position to the tablet through Bluetooth. Accelerometer and gyroscope readings were sampled at 600 hertz and combined with a complementary filter to reduce gyroscope drift and accelerometer noise.

The shoulder-wheel game asked children to steer a ship by aligning a target with a second ball controlled by the wheel’s angle. Calibration established the game’s zero position after accounting for the sensor’s physical mounting orientation. A sensor placed horizontally near the bottom of the wheel, for example, required an angular offset of minus 90 degrees relative to the game reference. During the final design stage, Treasure Hunter was also changed to accept partial movements rather than complete revolutions, allowing children with restricted ranges of motion to participate. The prototypes ran as Android applications on tablets and included separate child-facing and therapist-facing interfaces. Clinicians could adjust movement range, repetition targets, timing and other session parameters before play began.

Three structured focus groups were held for each game at successive development stages: after the initial concepts, after functional prototypes were operating, and after refined versions had been installed in the clinical setting. Sessions lasted approximately 45 to 60 minutes and covered therapeutic alignment, hardware setup, pacing, difficulty, feedback and data capture. The five staff members were selected purposively because they were directly involved in rehabilitation or device deployment. The two rehabilitation physicians and physical therapist attended all three sessions, while an informatics engineer joined discussions of database integration and the head of maintenance joined the installation and safety review. Transcripts and observational notes were used to identify requirements and agree on revisions, but no formal thematic-analysis framework or coding system was applied.

Several design changes followed directly from these discussions. Finger-ladder coverage was expanded, movement tracking was aligned more closely with therapeutic sequences, and time- and repetition-based goals were added. Both systems began storing session summaries in comma-separated values files for later use by therapists. Visual and auditory feedback were refined during final installation, and Hungry Hamster switched to rechargeable battery power following a safety recommendation. The researchers recommend integrating therapeutic gestures directly into the virtual environment, calibrating game mechanics to motor ability, presenting consistent feedback and using positive rewards without punishing failure. Simple narratives gave the repetitions a clear purpose: a hamster collects fruit while climbing a tree, while a ship gathers treasure at sea. Neither prototype included death, losing or other mechanics that might produce frustration.

Accessibility and pacing received particular attention. Clinicians recommended adjustable hardware, including height-adjustable tables secured with straps, so that children of different statures and wheelchair users could interact with the equipment. The study also recommends quieter screens, slow-moving decorative elements, strong but not overwhelming sounds, adjustable pacing and shorter or modular sessions. These provisions reflect the possibility that a child must simultaneously interpret the physical exercise and translate it into an on-screen objective, adding cognitive load to the motor task. Clinicians described two broad behavioral patterns without presenting them as validated diagnostic categories: some children were competitive and could work toward higher speed or repetition targets, while others became frustrated and needed fewer rigid goals, more time or rest periods. Therapist judgment was therefore treated as essential for responding to fatigue, stress, medication effects and changing willingness to participate.

Objective data capture emerged as another major recommendation because clinicians may otherwise describe the same performance differently. For the ladder, the systems recorded inter-step timing, total repetition time and maximum height; for the shoulder wheel, they recorded angular range, movement timing, repetition count and session duration. These measures were designed for monitoring activity and identifying changes across sessions, not for replacing validated clinical outcome assessments. The team also discussed timing every segment of a movement trajectory, because slower portions might identify restricted ranges that are hidden when only the final position is considered. Movement smoothness, posture control and other movement-quality measures were discussed in relation to prior literature, but the study did not quantitatively assess them. Automatic difficulty adjustment was also discussed, yet the implemented games relied on therapist-controlled configuration rather than a validated adaptive algorithm.

After the prototypes and staff training were complete, five children tested Hungry Hamster and 11 tested Treasure Hunter; two children used both systems. The research team had no access to ages, diagnoses, motor limitations, intellectual-disability status or functional classifications because of the center’s privacy requirements. Participants completed an exploratory 11-item Spanish-language questionnaire with five face-based response options, administered with help when needed. For Hungry Hamster, mean ratings across the first 10 items ranged from 3.8 to 4.6. When asked to compare exercise with and without the game, the former received a mean rating of 4.4, compared with 2.6 for conventional exercise. For Treasure Hunter, the corresponding means were 4.25 and 2.875. These are descriptive differences from small groups on an unvalidated ordinal questionnaire; they do not demonstrate that gamification caused better engagement or clinical benefit.

Several limitations sharply constrain the findings. The child sample was small, and convenience sampling plus absent demographic and clinical information prevented analysis by age, condition or functional profile. The questionnaire had no formal content validation, reliability assessment or construct-validity testing, and the Spanish-to-English reporting translation lacked independent forward-backward verification. Staff feedback came from a single center with particular workflows and resources, while testing lasted only through short prototype sessions rather than measuring long-term adherence, functional gains or comparative therapeutic effectiveness. The objective metrics were also not validated clinical outcomes. Nevertheless, the study offers a concrete development model: combine mechanical rehabilitation tools with direct sensor mapping, keep tasks and stories simple, let therapists control difficulty, design rewards around successful therapeutic movement, collect data that clinicians consider meaningful, and use age-accessible feedback methods. Larger longitudinal studies are needed before these prototype games can be presented as effective treatments or replacements for conventional rehabilitation.

Subject of Research: Clinician-informed design of serious games for children’s upper-limb rehabilitation

Article Title: Lessons learned from developing serious games for upper-limb rehabilitation in children

Article References: Sanchez, D. I., & Bourazeri, A. (2026). Lessons learned from developing serious games for upper-limb rehabilitation in children. Frontiers in Pediatrics, 14, Article 1904423. https://doi.org/10.3389/fped.2026.1904423

Image Credits: AI Generated

DOI: 10.3389/fped.2026.1904423

Keywords: rehabilitation games, upper-limb therapy, children, serious games, clinician-led design, motion sensors, gamified therapy, usability, adaptive difficulty, clinical data, Lessons, learned