The multicenter observational feasibility study, MyPal4Kids, was conducted from December 2020 to September 2022 at Saarland University (USAAR), Hannover Medical School (MHH; Germany), and University Hospital Brno (FNBRNO; Czech Republic). Recruitment closed on March 31, 2022 (last patient, first visit).

An observational study design was chosen to evaluate the feasibility of the platform in a real-world setting without randomization, enabling a naturalistic assessment of acceptability and engagement.

The study took place in pediatric oncology units at all 3 participating hospitals, which provide inpatient and outpatient care to about 290 new pediatric oncology patients annually, supported by multidisciplinary teams. Eligible participants were children and adolescents with cancer aged 6-17 years diagnosed with solid tumors or leukemia and treated at one of the sites within the previous 12 months.

Trained clinical staff recruited participants. Eligible patients and at least 1 parent or legal guardian were invited to participate in the 6-month study during an informational session. Written informed consent was obtained from patients and their parents. Study staff created user accounts, provided leaflets with QR codes for app installation and baseline questionnaires, and offered support as needed. Details of recruitment and eligibility are available in the protocol [45].

In addition to standard care, participants used the ePRO-based platform during their 6-month study participation. The platform had 3 components (Table 1) [50], which were available in German and Czech with a “read-aloud” functionality for younger users. A video demonstrating the MyPal platform and its apps is available online [53]. The platform was implemented through these 3 components and was offered for use within clinical workflows across user-selected devices. Participation was voluntary, and no mandatory actions or response pathways were established for HCPs. Data were securely transmitted and made available to HCPs to support clinical decision-making.

A participatory design approach guided the development of the platform with the aim to enhance its relevance, utility, and usability for the target population. This process included a literature review, prestudy focus groups [54], and structured discussions with patients, parents, and HCPs to identify needs, preferences, potential barriers, and desired functionalities. These activities served as a prestudy assessment to inform key design decisions and ensure that the platform was aligned with user requirements prior to implementation. Development followed an iterative feedback loop and included a 2-week pilot phase with children prior to study initiation to further refine the platform, identify technical issues, evaluate the local technical infrastructure, and obtain initial insights before the feasibility study commenced [50].

Participants were invited to use the platform and its apps for 6 months. Table 2 summarizes the study end points, outcome measures, and time points adapted from the published study protocol [45]. In addition to optional symptom reporting, patients and parents completed ePRO-based secondary end point measures via the apps. The questionnaires were selected based on validity, dissemination, digital availability, and author agreement (refer to Table 2 for end points and the assessment schedule and S5 in Multimedia Appendix 1 for questionnaire dimensions, rating scales, and scoring information).

Per protocol, the recruitment rate was defined as the proportion of recruited patients among eligible patients. The participation rate referred to recruited participants completing at least 70% of the 28 monthly questionnaires (before imputation). The attrition rate included patients who did not complete the 6-month study due to withdrawal or death. Reasons for missed assessments (multiple-choice and free-text responses) were documented when available.

Adherence was assessed by user group. For patients, adherence included the number of optional monthly symptom reports, diary entries, and in-game personalization activities. For parents, adherence included the number of monthly optional proxy symptom reports. For HCPs, adherence included the number of monthly treatment plans, reaction protocols, and proxy symptom reports. Engagement was defined as the frequency and consistency of platform use over time. Adherence referred to the completion of scheduled questionnaires, while participation described overall involvement in study activities.

Scoring and missing data handling followed the respective questionnaire guidelines (refer to Table 2). Additional details, including the full questionnaires, interview guides, coding schemes, paraphrased statements, and extended statistical outputs, are provided in Multimedia Appendix 1.

Focus groups were conducted at all three sites with (1) children and adolescents with cancer, their parents and (2) HCPs to obtain qualitative feedback on the platform (eg, experiences, challenges, barriers, and facilitators).

Semistructured interview guides, based on the mHealth App Usability Questionnaire [65], included prioritized, neutral, open-ended, unambiguous questions (refer to S3 in Multimedia Appendix 1 for patient and parent focus groups and S4 in Multimedia Appendix 1 for focus groups with HCPs). Patients and parents were approached post study. HCPs were divided into two subgroups: (1) project-internal HCPs who provided feedback during the platform’s development and (2) project-external HCPs who were involved only in the study.

Written informed consent was obtained before interviews (conducted in person or online and in the local language). Sessions lasted 45‐60 minutes and were moderated by trained local researchers. Audio recordings were transcribed verbatim, pseudonymized, and translated into English for analysis.

Statistical analyses were conducted in R (version 4.3.2, R Core Team, R Foundation for Statistical Computing) [66]. Descriptive statistics summarized recruitment and baseline characteristics (frequency, mean [SD], or median [IQR], as appropriate). Tests were chosen based on variable distribution. Two-sided P values <.05 were considered significant.

Missing data were imputed according to questionnaire guidelines. For symptom reports, the highest monthly symptom burden per patient was considered. Missing responses (≤2 per measure) were replaced by the nearest previous value. The target sample size was 100 participants [45]. To accommodate missing data and the small sample size, repeated measures (baseline and M1-M6) were grouped into 4 time points (baseline, M1-M2, M3-M4, and M5-M6), ensuring no significant differences between merged time points via the two 1-sided tests, with odds ratios calculated as effect sizes.

The aligned rank transformation (ARTool; developed by Kay et al. [67]) enabled a nonparametric 2-factor repeated measures ANOVA for the factors of time, group, and their interaction effects [68]. Participants were grouped by (1) birth sex (female or male), (2) age (6-9, 10-13, and 14-17 years), (3) diagnosis (leukemia or solid tumor), and (4) Karnofsky Performance Status (KPS) [69,70] scale (below or above the cohort median). Partial eta squared (η_p²) was classified as small (η_p²≥0.01), medium (η_p²≥0.06), and large effect size (η_p²≥0.16). Post hoc comparisons used Wilcoxon signed-rank tests (time points) and Mann-Whitney U tests (group comparisons), both with Holm-Bonferroni correction. For significant interactions, rank-transformed time-point differences between groups were compared using Mann-Whitney U tests with Holm-Bonferroni correction.

Spearman rank correlation (ρ) assessed monotonic associations between patient self-reports and parent proxy reports of symptom burden per month (no imputation). Effect sizes were classified as weak (|ρ|≥0.1), moderate (|ρ|≥0.3), and strong (|ρ|≥0.5).

Qualitative focus group data were analyzed in MAXQDA 2022 (VERBI Software) using deductive and inductive approaches [71]. Two researchers independently reviewed transcripts and developed a consensus coding system for thematic analysis. Statements were cross-coded when applicable (eg, positive, neutral, and negative) and paraphrased after joint review.

Missing data handling, subgroup analyses, and model specifications were performed as described in this section. No formal adjustment for site-level differences was performed because of the limited sample size at each site.

The MyPal4Kids study protocol has obtained ethical approval from the ethics committee competent for the Saarland University (Ärztekammer des Saarlandes; reference: Ha 23/20; March 16, 2020), the ethics committee of the Hannover Medical School (reference: 9095_BO_K_2020; February 12, 2020), and the ethics committee of the University of Brno (reference: 01‐1 20 220/EK; May 11, 2020). This study was performed in line with the principles of the Declaration of Helsinki. Ethical and General Data Protection Regulation (GDPR) compliance considerations ensured data safety for this vulnerable participant group [48], as reflected in the protocol through age-appropriate information sheets and informed consent forms [45]. Informed consent was obtained from all participants and from parents or legal guardians of children or adolescents.

Recruitment took place over 15 months (December 2020-March 2022), including a 6-month COVID-19 extension. A total of 83 participants were enrolled (refer to Figure 1).

Table 3 summarizes the characteristics of the enrolled patients. Birth sex distribution and treatment status were nearly balanced. Almost half of the patients were aged 14‐17 years. Solid tumors accounted for nearly half of the cases, with brain tumors listed separately due to their potential neurological effects [72]. Most patients were newly diagnosed (70/83, 84%) and 16% (13/83) of patients had relapsed.

MyPal4Kids assessed the feasibility and acceptability of an ePRO-based platform in pediatric oncology and palliative care. Patient engagement was strongly influenced by perceived system utility, linked to HCP feedback frequency and the game’s ability to sustain motivation. The core hypothesis that gamification would make self-reporting more engaging for young patients was partially supported: initial interest was high, but long-term engagement declined. These findings align with research emphasizing bidirectional communication in digital health [73-75], while also highlighting implementation challenges.

A key challenge was a disengagement cycle: fewer patient reports led to fewer platform reviews by HCPs, reducing feedback and further discouraging engagement. Similar patterns have been observed in other studies [76], likely amplified by the emotional burden on young patients with cancer and their families, along with competing HCP priorities. Usability did not limit patient engagement in our study but was more relevant for HCP adoption. These findings suggest that all groups may require additional support, such as proactive feedback loops and tighter integration into clinical workflows and systems, to maintain engagement. Engagement decreased over time, which represents an important finding of this study. This pattern was analyzed using descriptive longitudinal summaries rather than formal subgroup modeling. Engagement reflects a behavioral usage process characterized by heterogeneous exposure durations, irregular participation intervals, and varying initiation and discontinuation points across participants. This differs from structured repeated-measures clinical outcome assessments collected at predefined time points, which are more suitable for subgroup comparisons within the feasibility framework. To complement the quantitative description, we also further provide stratified descriptive observations (eg, age-related differences in usability and engagement patterns) alongside qualitative findings that help contextualize these trends, including differences in motivation, disease status, and HCP feedback.

Potential explanations for the observed decline include limited novelty effects, insufficient personalization of gamification elements, and varying needs across age groups. Additionally, the degree of feedback from HCPs may have influenced sustained use. These findings highlight the need for adaptive and user-centered design strategies to maintain long-term engagement. In addition to the COVID-19 pandemic, institutional factors such as varying levels of digital infrastructure, staffing resources, and staff engagement may have influenced implementation processes and platform use across study sites.

Similar engagement trajectories have been reported in pediatric digital health and ePRO-based interventions, where engagement is typically high at initiation but declines over time despite generally acceptable usability. Prior pediatric ePRO studies have similarly described decreases in reporting frequency in longitudinal implementations [77-79], as well as in broader mHealth and game-based interventions [80,81]. These findings are consistent with evidence that gamification strategies often produce short-term engagement effects but are insufficient to sustain long-term use [30]. In addition, the effectiveness of social or competitive design elements may depend on a sufficiently large and continuously active user base, which is often lacking in pediatric feasibility settings [82]. Together, these observations suggest that sustained engagement requires more than initial usability and gamification and depends on continued perceived relevance, feedback mechanisms, and integration into clinical workflows.

Importantly, this study was designed as a feasibility study and was not intended to evaluate the effectiveness of the intervention. Therefore, any observed changes in ePROs should be interpreted descriptively and not as evidence of causal effects. The platform proved feasible for tracking individual symptom burden over time, contingent on patient engagement in reporting symptoms. Clinicians could benefit from reviewing individual symptom trajectories. Statistical analyses showed significant main and interaction effects of time and group. Larger datasets could enable more nuanced comparisons, enhancing the platform’s potential for research applications, such as comparing treatments in terms of effectiveness or side effects. This dual capability, supporting individualized patient care while generating population-level insights, underscores the platform’s potential for both clinical practice and research. While symptom burden was the primary focus, secondary measures like children’s QoL also showed improvement from baseline until the end of the study. Interpretation of these findings should remain cautious, as broader pandemic-related disruptions may have influenced reporting behaviors, engagement patterns, and ePROs during the study period [83-85].

Underlying factors may include age, digital literacy, emotional strain, and institutional support for digital health tools. Assessed symptoms largely reflected expected chemotherapy side effects, about which patients and parents had prior information. Consequently, HCPs sometimes gave limited direct feedback. Still, reports of symptoms such as tingling or numbness in the extremities are clinically relevant, potentially guiding treatment adjustments of medication doses. Our findings reinforce concerns that insufficiently integrated digital health tools can increase rather than reduce workload for HCPs, particularly when clear integration pathways into routine clinical workflows and responsibilities are lacking [75,86,87]. In the present study, the platform was deliberately not embedded into mandatory clinical workflows, which may have contributed to this effect. Existing commercial clinical systems offer limited interoperability with digital health platforms, including electronic health record integration, application programming interfaces and authentication infrastructures (eg, different user accounts and password systems). This complicates seamless integration into routine clinical workflows and requires the use of parallel systems. Dedicated staff to review digital health data could alleviate this burden, an approach supported in the literature [12]. Timely symptom data are crucial for responsive care, particularly between infrequent visits [88]. The observed decline in symptom reporting mirrors similar studies, possibly due to symptom relief or engagement fatigue. While gamification sparked initial interest, it alone was insufficient to sustain engagement, especially in older children. The game’s design was constrained by requirements for nonviolence, gender neutrality, and broad age group appeal. While these specifications ensured accessibility, they may have resulted in an overly cautious design. Concerns about overuse did not materialize. Established gamification elements such as rewards and competitive features were less effective without a critical mass of simultaneous active users. Perspective differences between internal and external HCPs should be considered. Internal HCPs, involved in platform development, focused on long-term potential, while external HCPs emphasized practical challenges like workflow integration. Both groups agreed that technology should complement, not replace, direct care [89].

Interdisciplinary collaboration and testing across multiple clinical sites strengthened the study. Patient, parent, and HCP involvement in the design and evaluation enhanced the relevance of the findings. Ethical considerations were central to platform development and study implementation. Integration of ePROs into a gamified mobile app for young patients with cancer was innovative, aligning with evidence that children prefer visually engaging digital measures with short recall periods [90]. A strength of the presented platform is its flexibility for various ePROs and gamification strategies, enabling implementation across varied contexts.

The study period overlapped with the COVID-19 pandemic, which likely affected recruitment, clinical workflows, and sustained platform engagement. Reduced face-to-face interactions, restrictions during clinical visits, and increased reliance on digital communication may have influenced onboarding, participant retention, and reporting behavior. Prolonged exposure to digital communication and online platforms during the pandemic may also have contributed to digital fatigue and declining engagement over time. Similar pandemic-related disruptions affecting recruitment, ePRO-based research, and health-related QoL assessments have been described in pediatric and digital health research [83,85,91].

Further limitations include discrepancies between child and parent symptom reports, possibly due to recall issues or differing perceptions. While proxy reports can increase data completeness, they may introduce bias, highlighting the importance of careful interpretation and, where possible, cross-checking [92]. Some parents may have prioritized immediate care over proxy reports. Dual reporting was optional, limiting analyses to overlapping cases. Still, at certain time points, some symptoms demonstrated strong parent-child correlations, supporting limited proxy validity. MyPal was not designed as an emergency tool, as stated in the consent. Legal and ethical concerns, such as liability for missed reports, implicated that HCPs were not expected to monitor reports constantly. This limited the platform’s immediate clinical utility but matched its intended use as a supplemental communication channel. The small sample size limits the robustness and generalizability of the findings. Given the limited sample size and reduced numbers in subgroup analyses, these findings should be interpreted with caution. The results are exploratory and intended to inform future studies rather than provide definitive conclusions. The 6-month study period may not reflect long-term patterns. Given the drop in engagement, longer studies may require design refinements. The sample was limited in terms of diversity (eg, age and disease types). Selection bias is possible, as families with higher digital literacy or strong clinical engagement may have been more likely to participate. The overrepresentation of adolescents (aged 14‐17 years) does not reflect the general pediatric cancer population [93]. Younger children younger than 4 years have the highest cancer incidence but were excluded due to literacy barriers. Treatment plan creation in the platform was optional for HCPs, limiting analysis of symptom-treatment correlations. Recruitment and HCP availability were likely affected by the pandemic, though the impacts are hard to quantify. Emotional burden during early illness phases also influenced recruitment. Subjective feedback from participants within the clinical setting may have shaped both recruitment and engagement. Measures such as flexible baseline and monthly assessment windows, as well as offline functionality, were implemented to support compliance, though their success was limited. Secondary end point measures (eg, patients’ and parents’ QoL, care satisfaction, and family impact) were not available to HCPs during the study. Access to these data could increase HCP engagement.

Taken together, these factors may limit the generalizability of the findings beyond European pediatric oncology settings, particularly in contexts with different health care infrastructures, cultural expectations, and levels of digital literacy.

Although more digital health solutions are emerging for specific illness phases, creating comprehensive platforms that address diverse individual needs remains challenging. Concerns persist about using research-oriented ePRO measures in clinical decision-making [8]. Pediatric-specific clinical ePRO guidelines are lacking [94], underscoring the need for evidence to support standardization and broader clinical adoption. Future versions should consider structured HCP training and targeted family support. Embedding ePROs into standard care rather than offering them as optional tools may help sustain use. Refinements should reflect user feedback and improve age-specific adaptability to enable scalability across clinical settings. Hybrid models, combining automated feedback with periodic HCP review, may boost engagement and long-term utility. Digital health solutions may not suit all patients; tailoring them to specific populations is key. Longitudinal randomized controlled trials are needed to assess effectiveness. Future research should explore the use of the platform in different patient populations, such as long-term survivors, who may demonstrate different supportive care needs and engagement patterns. In addition, studies conducted in postpandemic settings may help clarify the extent to which engagement patterns, recruitment dynamics, and sustained platform use were influenced by pandemic-related contextual factors. Evaluating the platform in larger cohorts and in more controlled study settings will be important next steps.

In conclusion, integrating digital health tools into pediatric oncology and palliative care requires both technological innovation and thoughtful workflow integration. Sustaining participation and utility demands consistent HCP engagement and compelling gamification. Large-scale implementation will require dedicated resources, including staff time for data monitoring, integration into existing hospital IT infrastructures and application programming interfaces, authentication systems, and sustainable allocation of clinical responsibility for digital symptom review. Our findings highlight key areas for refinement to improve ePRO-based approaches before clinical use is feasible.