We used purposive, maximum-variation sampling to capture a diverse range of experiences and perspectives. The inclusion criteria included (1) parent or legal guardian of a child aged 6‐10 years diagnosed with amblyopia, (2) pediatric ophthalmologist with ≥3 years of clinical experience, or (3) child meeting diagnostic criteria who could assent with parental help. The exclusion criteria included inability to communicate in Mandarin or refusal to be recorded. Parents were approached during clinic visits; clinicians were invited by email; children were recruited via clinician referral. In total, 8 ophthalmologists, 22 parents (8 families), and 6 children participated. The sample size was guided by information power, and theoretical saturation was reached by the sixth parent interview, as no new codes or themes emerged, indicating sufficient coverage of core concepts.

Semistructured interview guides were pilot-tested with 2 parents and 1 clinician, then refined for clarity and relevance. The interview topics covered barriers to occlusion therapy, child motivation, game preferences, and parental support. The interview outlines for parents and children are available in Multimedia Appendices 1 and 2.

All interviews were conducted by 2 trained researchers (a female design graduate student and a male MSc research assistant in ophthalmology) in private clinic rooms (for parents or clinicians) or a quiet play area (for children). Each interview lasted 40‐60 minutes, was audio-recorded, transcribed verbatim, and securely backed up on encrypted drives. Member-checking was used for credibility: summary transcripts were sent to 4 participants to confirm the accuracy and resonance of identified themes.

Transcripts and questionnaires were analyzed using inductive thematic analysis in NVivo 14: familiarization with data, initial coding (open coding), theme development (axial and selective coding), theme review, theme definition/naming, and reporting [55]. At least 2 coders independently performed each coding round; disagreements were discussed with a third coder until consensus was reached, resulting in high intercoder reliability (κ=0.82). At least 2 anonymized, supportive quotations per theme are provided in the Results section for confirmability.

Reflexivity and trustworthiness were prioritized throughout analysis. Researchers maintained a reflexive diary to document positionality, preconceptions, and their influence on data interpretation. Confirmability was established through audit trails and member checking. Dependability was ensured via transparent documentation of all analytic steps and coder discussions. Credibility was strengthened by participant feedback; thick description; and triangulation between parent, child, and clinician interviews. Transferability is enhanced by detailed reporting of sample characteristics and recruitment context, enabling readers to assess relevance to other settings.

An expert panel was assembled using purposive sampling across hospital, academic, and design networks to capture diverse knowledge relevant to pediatric amblyopia treatment and gamified intervention development. A total of 6 experts were recruited, including 3 pediatric ophthalmologists (each with at least 5 years of clinical experience) and 3 game designers (each with a minimum of 5 years of experience in serious games or health-related gamification). The inclusion criteria comprised verified domain expertise and willingness to engage in 2 rounds of structured consultation.

In Round 1, each expert participated in a 45‐ to 60-minute Zoom interview focused on operationalizing the 5 dimensions of cognitive appraisal theory (relevance, valence, likelihood, agency, coping potential) within the concrete mechanics, narrative structures, feedback modalities, and visual styles suitable for pediatric rehabilitation. Interview questions, refined following pilot feedback, are available in Multimedia Appendix 3. All interviews were audio-recorded, transcribed verbatim, and thematically coded by 2 independent coders in NVivo 14 following an inductive-deductive hybrid approach; intercoder reliability for theme extraction was κ=0.79. The team used reflexive diaries to minimize bias, ensuring that both clinical and design perspectives were adequately represented. Coded insights from all experts were synthesized into a design-decision matrix that mapped cognitive appraisal dimensions to actionable game features. This matrix guided ideation of low-fidelity paper sketches, followed by a high-fidelity Figma digital prototype (Prototype 1) that reflected prioritized features and narrative components identified in Round 1.

In Round 2, experts were provided asynchronous access to Prototype 1 and asked to rate each proposed feature on a 7-point Likert scale for relevance and feasibility. Consensus was defined a priori as at least 80% agreement per domain and item. All items met or exceeded this threshold after minor revision based on expert feedback, resulting in a finalized intervention prototype (Prototype 2) used in the subsequent RCT. Reflexivity was maintained during consensus scoring, and disagreements were reconciled through written clarification and reevaluation to ensure the transferability of core design principles to clinical practice and child end-user contexts.

Children aged 7-10 years with refractive, strabismic, or mixed amblyopia were recruited via school outreach and clinic referrals. The inclusion criteria were a confirmed diagnosis, ability to follow the training with parental support, and access to a mobile device at home. The exclusion criteria included severe amblyopia (best corrected visual acuity≤0.2), other ocular or neurological comorbidities, and prior exposure to digital amblyopia therapy. Based on pilot data (expected mean difference 1.5 on MMAS-8, SD 1.2; Cohen d=1.25), 15 participants per arm were required to achieve 90% power at α=.05 (2-tailed). We therefore enrolled 40 participants to mitigate risk of loss to follow-up and randomized after consent or assent.

An independent statistician generated a computer-based permuted-block sequence (block size=4). Allocation was concealed using sequentially numbered, opaque, sealed envelopes. Outcome assessors and the statistician were blinded to group assignment.

After written parental consent and child assent, participants completed a baseline enrollment visit during which we verified eligibility and collected demographic variables on a standardized case report form: age (in years), sex (male/female), and school grade (3rd-5th), as well as device availability for home use; demographic sources were parent report at enrollment using a structured checklist. Participants were then randomized to the intervention (our gamified app, “Find You! Cure My Animal Friends”) or the control (a widely used commercial visual-training app) and received a standardized orientation, including a brief instructional video for parents and a printed quick-start guide covering daily use, patching alongside play as clinically indicated, and troubleshooting. Families were instructed to complete home-based training for 20 minutes per day over 4 consecutive weeks; no additional clinical care was altered for either group during the study period. To support consistency, parents were asked to supervise session start/stop and encourage completion at roughly the same time each day.

During data collection, 6 participants (3 per arm) withdrew (reasons reported to the study team: lack of reliable home internet, scheduling conflicts, or unrelated illness). Data from these participants were not used in the analysis. The final analytic sample was N=34 (intervention n=18; control n=16).

The primary outcome was treatment adherence, assessed with an adapted MMAS-8 (licensed from Morisky LLC; see Multimedia Appendix 4 for item wording and scoring) [56,57]. Within the adaptation, medication-specific terms were reworded to “vision training” and “eye patch” to ensure contextual relevance for pediatric amblyopia. Comprehensibility and content validity were confirmed via cognitive interviews with 5 parent-child dyads prior to trial launch; pilot administration with n=20 yielded Cronbach α=.83, supporting internal consistency. The MMAS-8 was administered at the end of week 4 with parental assistance, and total scores were classified into low, medium, or high adherence according to developer guidance.

Secondary outcomes included the UEQ (26 items across 6 dimensions, see Multimedia Appendix 5) and postintervention semistructured interviews with children and parents (20‐40 min, conducted via Zoom, see Multimedia Appendix 6). These interviews explored emotional and behavioral engagement with the intervention, thematically mapped to cognitive appraisal theory dimensions (relevance, valence, likelihood, agency, and coping potential).

All interviews were audio-recorded, transcribed verbatim, and analyzed via dual-coder thematic analysis in accordance with COREQ standards. Saturation was reached when no new themes or codes emerged during review, and at least 2 anonymized supporting quotations per theme are reported for confirmability. Intercoder reliability was assessed, and transcript review was supplemented by member-checking for credibility. Reflexivity was maintained by both coders using analytic memos to document positionality and potential biases.

Table 2 shows a quantitative summary of stakeholder interview themes and subthemes from Stage 1 qualitative needs assessment in a 3-stage mixed methods study developing a gamified mobile intervention for pediatric amblyopia rehabilitation. Two overarching themes—Desires and Preferences—were derived through inductive thematic analysis with dual coding in NVivo 14 under COREQ-consistent procedures, and each subtheme is reported with participant proportion and an example code. Within Desires, adherence-relevant subthemes included sustained engagement (75%), enhancing the appeal of the eye patch (50%), rewards and immediate feedback (65%), and emotion regulation concerning feelings about being “in treatment,” “being ill,” and “wearing an eye patch” (95%). Within Preferences, children favored colorful collectibles (60%) and hands-on activities such as puzzles and drawing (45%). Interviews used purposive, maximum-variation sampling and included 8 pediatric ophthalmologists, 22 parents (8 families), and 6 children, conducted in clinic and child-friendly settings at the Eye & ENT Hospital of Fudan University.

The stakeholder interviews revealed a strong preference among parents and children for games that sustain attention over time. Key design motivators included continuous storylines; meaningful role-play; and clear, simple instructions—preferred over repetitive, single-player mini-games. These findings align with the high endorsement of narrative elements (75%) as reported in Table 2. Quantitative summary of theme frequencies, coded proportions, and representative quotations are presented to illustrate key patterns and their implications for therapy design.

Families described both physical discomfort and peer-related embarrassment as major barriers to consistent patching (50%). Suggestions included attractive, customizable patterns and integrating patch-related actions with meaningful in-game events so that wearing the patch feels purposeful rather than punitive.

Parents reported that immediate, varied rewards are more effective than vague, delayed promises in sustaining daily participation (65%). Several children also mentioned striving for a high score as intrinsically motivating, while parents cautioned that repeating the same reward can lead to boredom.

Interviews recurrently pointed to reducing stigma, fear, and frustration associated with treatment routines and visible patching (95% of parents). Typical comments included “If the patch could be part of the story—like a special pass to enter the game world—my child wouldn’t feel ‘sick’; she’d feel proud to wear it.” Parents also highlighted the need for gentle hints and affirming messages when children get stuck, and for empathetic narratives that normalize help-seeking and provide language to address teasing at school.

Mapped to cognitive appraisal theory, these requests correspond to valence (positive affect and feedback), agency (controllability through role-play), coping potential (scaffolding and hints), relevance (linking patch use to core game goals), and likelihood (clear goals and immediate feedback that reduce uncertainty).

Under preferences, children spontaneously discussed liking bright, saturated colors (eg, pink, orange, blue), collectible items (stickers, stationery), and hands-on activities (puzzles, building, sketching). Many favored comic-style narratives and “becoming a character,” further supporting story-driven, role-based mechanics.

Synthesizing these findings, we propose 4 actionable recommendations for pediatric amblyopia training: (1) deliver quick, varied positive feedback to sustain engagement; (2) personalize content to children’s preferences and abilities; (3) embed rich storyworlds (animal companions, collectible artifacts) to deepen immersion; and (4) incorporate emotion-regulation affordances—empathetic cues, reframing of patching, and progressive challenges—to mitigate negative feelings surrounding treatment and illness and thereby strengthen adherence.

Table 3 shows quantitative support from expert interviews in Stage 2, mapping the 5 cognitive appraisal dimensions to occlusion-therapy optimization and rehabilitation game-design principles for a gamified pediatric amblyopia intervention. This table summarizes expert-endorsed design mappings across Relevance, Valence, Likelihood, Agency, and Coping potential, reporting for each dimension the corresponding occlusion-therapy focus, game-design guideline, proportion of expert agreement, and an illustrative solution to support adherence. The expert process comprised 2 interview rounds with 6 experts (3 pediatric ophthalmologists and 3 game designers; ≥5 years’ experience), Zoom-based sessions (45-60 min), thematic coding with intercoder agreement κ=0.79, and an a priori consensus threshold of ≥80% on relevance and feasibility, informing Prototype 2 used in the RCT. Collectively, these principles specify how occlusion-therapy behaviors and game mechanics should be integrated to enhance adherence in pediatric amblyopia care.

Experts recommended treating the eye patch as a meaningful artifact in the game world (eg, a narrative “pass” or “gift” from animal companions) to reframe patching as purposeful rather than punitive (Relevance). Progress in the game should depend visibly on correct patch use (Valence/Likelihood), and a short tutorial animation should demonstrate proper wear to reduce misuse and frustration (Coping potential).

Rehabilitation game mechanics were mapped to the following appraisal dimensions:

Synthesizing these inputs, we adopted interactive storytelling as the primary format. Players act as an intern doctor who completes daily treatment tasks for animal patients (visual-training tasks embedded). The game includes 3 difficulty levels selectable by the child, collectible illustrations and achievements as incentives, and a child-friendly cartoon style. These principles were consolidated into a design-decision matrix and implemented in the intervention app used in Stage 3 (see Figures 1,2; Table 4).

The gamified mobile app “Find You! Cure My Animal Friends” developed for this study was designed to enhance treatment adherence in children with amblyopia by integrating cognitive appraisal theory. It provides a series of visually stimulating activities that are specifically crafted to improve visual acuity and encourage regular use, as shown in Figure 1. The main game interface and level interface are shown in Figure 2.

Table 4 shows baseline characteristics of participants by randomized arm in the Stage 3 parallel-group RCT evaluating a gamified mobile app versus a commercial visual-training app for treatment adherence in pediatric amblyopia. Children aged 7-10 years were recruited via school outreach and clinic referrals in Shanghai, China, yielding an analytic sample of N=34 (intervention n=18; control n=16) after attrition. Variables include age distribution (7-10 y) and sex, with no statistically significant between-group differences at baseline.

The intervention and control groups yielded W=0.946 (P=.37) and W=0.963 (P=.71), respectively, indicating no evidence against normality. Normality assessment informed the use of independent-samples t tests alongside nonparametric tests in SPSS 26.0 with 2-tailed α=.05, as prespecified in the statistical analysis plan.

After conducting an independent samples t test, a significant difference was found in the adherence scores on the MMAS-8 scale between the experimental group that used the amblyopia training game (t₃₂=3.896; P<.001). The experimental group (N=18) had a mean adherence score of 6.56 (SD 1.06, SE 0.25, 95% CI 6.03-7.08), while the control group (N=16) had a mean score of 5.01 (SD 1.22, SE 0.31, 95% CI 4.36-5.66). The results indicated that the experimental group had significantly higher adherence scores than the control group.

Significant behavioral differences were found when comparing patient adherence to an amblyopia training game (experimental group) with other games on the market (control group). It was revealed that 61.11% (11/18) of patients in the experimental group showed high adherence, whereas only 18.75% (3/16) in the control group achieved the same level of adherence(P=.012). Additionally, the experimental group demonstrated a significantly lower percentage of patients with low adherence: 38.89% (7/18) in the experimental group versus 81.25% (13/16) in the control group(P=.012).

Table 5 shows the comparison of adherence issues between the intervention group and the control group. We provided the percentage of subjects who scored 0 for Q1-Q7, along with the mean score for Q8. The results revealed higher percentages in the control group for Q2, Q5, and Q7, which align with the significant differences observed in our statistical analyses. Significant differences emerged in some questions, notably “Have you missed any doses in the past two weeks?” (Q2: 11.11% vs 56.25%, t₃₂=3.03; P<.001), indicating a higher compliance in the experimental group. Similarly, for “Did you find it challenging to adhere to the treatment program?” (Q7: 11.11% vs 50.00%, t₃₂=2.59; P=.02), the experimental group showed significantly better adherence. No significant differences were observed between the experimental and control groups for questions about forgetting to take medication (Q1: 22.22% vs 31.25%, t₃₂=0.58; P=.57), adjusting medication without informing the doctor (Q3: 11.11% vs 12.50%, t₃₂=0.12; P=.90), and whether they took their medication the previous day (Q5: 11.11% vs 50.00%, t₃₂=2.59; P=.02).

Table 6 shows the UEQ scores by randomized group at the 4-week assessment in the Stage 3 parallel-group RCT evaluating a gamified mobile app versus a commercial visual-training app for pediatric amblyopia rehabilitation (children aged 7-10 y, Shanghai, China, March-December 2024). The table reports means, standard deviations, sample sizes, and 95% CI for 6 UEQ dimensions (Attractiveness, Understandability, Efficiency, Reliability, Facilitation, and Novelty) in the intervention (n=18) and control (n=16) arms as a prespecified secondary outcome. The intervention outperformed the control on Attractiveness (2.14 vs 0.39), Understandability (1.88 vs 0.19), Facilitation (1.92 vs 1.14), and Novelty (2.13 vs 1.09), while the control showed a slight advantage on Efficiency (1.20 vs 1.08); Reliability was comparable (1.35 vs 1.30)

Table 7 shows postintervention individual interview results by randomized group mapping children’s experiences to cognitive appraisal dimensions in a 3-stage mixed methods study of a gamified mobile intervention for pediatric amblyopia rehabilitation (Stage 3 RCT, Shanghai, China, children aged 7-10 y). Interviews with children and parents were conducted via Zoom (20-40 min) after the 4-week intervention and thematically analyzed with dual-coder procedures consistent with COREQ, with codes organized under 3 themes: rehabilitation training experience, attitudes toward rehabilitation, and attitudes toward wearing eye patches. Codes are presented for the intervention and control groups and annotated with the corresponding appraisal dimension (relevance, valence, likelihood, agency, and coping potential) to show directional changes in perceived engagement and coping during treatment.

The study found that children in the experimental group had a positive reaction to the gamified rehabilitation training. This positive response may be attributed to the game design being consistent with cognitive appraisal theory. Children particularly enjoyed storytelling, role-playing, interacting with in-game animal characters, and collecting icons. These elements increased their perceptions of goal relevance and congruence, as well as their assessments of the likelihood of successfully completing the game. Furthermore, enabling the children to actively engage in the game and interact with the surroundings enhanced their sense of control and ability to cope, which could account for the experimental group’s greater participation in the game and adherence to rehabilitation training. The experiences and attitudes of the children in the control group differed significantly, as they mostly expressed disinterest in the game content or found the images unattractive. However, even in the control group, children were willing to play the game repeatedly to achieve a high score, indicating that perceptions of likelihood and coping potential can still be improved with appropriate incentives. Furthermore, the experimental group’s children exhibited a higher level of positive acceptance toward the specially patterned eye patches, which provides additional evidence that increasing goal relevance and consistency can enhance treatment adherence. Conversely, the control group’s children were less receptive to eye patches and needed parental intervention.