The open trial, conceptualized by 3 authors (HT, KS, and SM), used a mixed-methods design, including quantitative analysis of anxiety symptoms and quality of life outcomes, intervention use, and qualitative analysis of participant feedback.

A total of 24 young people aged between 10 and 17 years were recruited from a tertiary children’s hospital in Auckland, New Zealand, between October 2018 and May 2020. Eligible participants had any type of long-term physical condition lasting for longer than 3 months (eg, asthma, diabetes, cancer, and cystic fibrosis) and measurable levels of anxiety (eg, specific phobia, generalized anxiety, and nonspecific anxiety) and may or may not have had comorbid mental health conditions. Eligible participants were of any ethnicity, intellectually and physically able to use the intervention, and understood enough English to play the game and provide informed consent or assent with paired parental consent if they were aged <16 years. Participants who did not meet all these criteria and those who had recently undertaken or were undertaking CBT or other forms of psychotherapy, biofeedback therapy, or pharmacotherapy with anxiolytic medication within the past 6 months were excluded because the effect of those therapies could confound the impact of the intervention.

Starship Rescue is a game-based eHealth intervention based on the story of a space hospital (starship) that gets caught up in a vortex of anxiety. The narrative is a new captain whose mission is to help find the lost bravery stars and restart its engine. Purposely designed to harness the correlation between shorter duration of use and outcomes [41], it includes 5 modules, each taking between 15 and 30 minutes to complete. Module 1 introduces players to anxiety and its origins and features, module 2 focuses on beating anxiety using their bodies, for example, via deep breathing and progressive muscle relaxation, module 3 teaches players how to discern between helpful and unhelpful thoughts and to prioritize the former, module 4 introduces problem solving and graded exposure to address smaller and bigger forms of worry or anxiety, and module 5 is a final quiz to consolidate the learning from previous modules. On completion of the game, players are emailed a summary of learned techniques in the form of a stay cool capsule. The intervention is provided on a tablet synced with a commercially available, wrist-based Scosche Rhythm Plus heart rate monitor which is accessed during biofeedback-based relaxation exercises. Starship Rescue is underpinned by the principles of (1) CBT, (2) biofeedback, (3) learning theory, and (4) game player taxonomy. Although most knowledge and skills are gained within modules, users must also leave the game and practice overcoming a named worry or anxiety in the real world between the fourth and fifth modules to complete the intervention. No external therapist support is required. Parental involvement has previously been shown to aid the successful completion of eHealth interventions, learning, and application of skills and systemic risk factors associated with the maintenance of childhood anxiety [42]. During the use of Starship Rescue, parents help their child choose a real-life reward to receive on completion of the intervention, validate the achievement of an out-of-game task in the fourth module by entering a four-digit code so that they can proceed to the final module, and are emailed a summary of the key learning points if their child does not have an email address. Participants were loaned a tablet with Starship Rescue installed on it and encouraged to complete all modules at home or in the hospital within 8 weeks. If they requested additional time at this stage, it was provided. Player data were saved within the game, allowing participants to pause and resume the game when participants wished to do so. Some data (eg, module completion and time taken) were manually pushed by a member of the research team from the game to an administrator email address at the end of the game. Further details of the modules and theoretical underpinnings are provided in Multimedia Appendix 1 [43-54]. Illustrative images are presented in Figure 1.

The primary outcomes of the open trial were evaluated as follows: (1) acceptability of the prototype intervention (ie, is the content and format acceptable to users?) was quantitatively assessed via user ratings of overall acceptability and helpfulness on scales from 0 to 10 and qualitatively assessed via feedback during semistructured interviews following completion of the game; (2) usability of the intervention (ie, is it usable?) was quantitatively assessed using the System Usability Scale (SUS) [55], time taken to complete the game and module completion and qualitatively assessed via feedback during semistructured interviews following completion of the game; and (3) clinical effectiveness (ie, does it reduce anxiety and related issues?) was assessed by measuring changes over time in the Generalized Anxiety Disorder, 7-item (GAD-7) [56], Spence Children’s Anxiety Scale (SCAS) [57], a Likert scale of anxiety embedded in the game, and the Pediatric Quality of Life Inventory (PedsQL) [58], as outlined in the schedule below (Table 1). The GAD-7 is a brief (7-item), self-reported scale for measuring anxiety in people aged ≥13 years. It has a sensitivity of 89% for generalized anxiety disorder, and scores of 5, 10, and 15 out of a possible 21 points indicate mild, moderate, and severe anxiety levels, respectively. The SCAS is a well-validated but longer (46-item) self-reported scale measuring child anxiety with sound psychometric properties with internal consistency reported at 0.92 for the total child score. It contains 6 subscales for panic or agoraphobia, social phobia, separation anxiety, obsessions or compulsions, fear of physical injury, and generalized anxiety. Likert scales (linear scales with items or numbers, eg, 0-10) have been shown to be useful for monitoring changes in anxiety [59] but are limited by user avoidance of extreme ratings [60]. Visual analog scales (involving images such as faces of differing sizes or nature) have also been shown to be useful for rating anxiety [61] with superior measurement qualities [62]. There is some disagreement about which type of scale is better for use with children [63,64]. Repeated brief evaluation of anxiety using simple measures such as Likert visual analog scales within a game, such as Starship Rescue, can be considered a form of ecological momentary assessment (EMA). EMA has been demonstrated to be useful for providing a richer picture of how behavioral changes occur over time [65]. To embed a repeated and accessible EMA within Starship Rescue, a combined Likert visual analog scale was developed that includes the face of an anxiety monster (as shown in Figure 1). This face can be moved with a slider to the desired point from the left (also marked 0) side of the screen to the right (also marked 10) and enlarges as this occurs. The PedsQL is a well-validated, 23-item self-report or parent-report scale measuring the quality of life. It has good internal consistency (0.88 for total scale), validity, and acceptability. It reliably distinguishes between healthy children and those with acute or long-term physical conditions.

Quantitative data were analyzed by our biostatistician (CF) using Excel (version 16, Microsoft Inc) and SPSS (version 25, IBM Corp). Analyses included basic descriptive statistics (eg, number of sessions completed, number of times device accessed, duration of use, changes in anxiety score, and demographic characteristics of the sample). McNemar chi-square tests and one-tailed t tests were used to assess the statistical significance of changes in anxiety scores over time. P values of <.05 were taken to indicate statistical significance, and 95% CIs were used to establish the extent of any difference between pre- and postmeasures. A sample size of at least 20 was calculated a priori to detect changes within the study group with effect sizes of 0.65 or more as statistically significant (α=.05 with 80% power). Data from this trial will be used to inform power calculations for a more definitive RCT. An intention-to-treat analysis was used with missing data managed using the last observation carried forward method. Qualitative data were manually analyzed using a general inductive approach [66] with collated text independently analyzed by 2 researchers (HT and HK) and any discrepancies addressed by consensus.

This study received ethics approval from the New Zealand Health and Disability Ethics Committee (HDEC, 16/CEN/136) on September 30, 2016. The lower age limit for participation was initially set at 12 years but later extended down to 10 years following a period of slow recruitment. Invitations to participate in the study were forwarded to potential participants through their own clinicians to minimize coercion using a direct approach. Verbal and written consent was obtained directly for those aged >16 years and via their parents with paired participant assent for those aged <16 years. Participants were free to discontinue engagement at any stage without consequence, and this was made clear to them. Although plans were made for any unanticipated distress occurring during participation to be managed by immediate referral to the hospital-based pediatric consultation-liaison mental health team, of which the lead author (HT) is a team member, this never occurred. Data were securely stored on a department server and kept securely for 10 years (or 10 years following younger participants’ 16th birthday) according to the ethics committee requirements.

A total of 15 participants aged between 8 and 16 years, of mixed gender (10 males and 5 females) and with different long-term physical conditions (cancer, asthma, bronchiectasis, cystic fibrosis, Alport syndrome, and others) provided feedback, 2 of which on multiple occasions. User feedback was incorporated to address technical issues, make instructions clearer, and develop the game’s look and feel. Examples of user feedback during the first sprint and the use of this feedback are provided in Textbox 1. By the end of the third cycle, there were sufficiently minimal technical issues and common concerns to proceed with the open trial.

Our findings provide preliminary evidence that Starship Rescue is an acceptable, usable, and effective new eHealth intervention for treating anxiety and improving quality of life in children and young people with long-term physical conditions. They also confirmed the feasibility of undertaking a larger RCT to confirm these findings. Starship Rescue appears to have comparable effectiveness (Cohen d=0.6) with existing eHealth interventions designed to address anxiety in children without long-term physical conditions such as BRAVE online (Cohen d=0.76) [28]. More than 85% (21/24) of our sample demonstrated clinical improvement immediately following intervention, and most maintained this benefit at the 3-month follow-up. Starship Rescue also appears to be more effective than Web-MAP, the only CBT-based, chronic pain–focused, web-based intervention identified by a recent Cochrane review for treating anxiety in children aged 11-17 years with long-term physical conditions (Cohen d=0.53) [67].

We believe that positive design features of Starship Rescue include a smaller number of modules (n=5) than other eHealth interventions for anxiety such as BRAVE online (n=16) and Cool Kids online (n=8), the reduction and tunneling of CBT content to improve adolescent engagement [68], and the inbuilt ability to repeat key skills to achieve mastery [69]. The lack of therapist support makes Starship Rescue more cost-effective than existing eHealth interventions for childhood anxiety. Despite the concern of other researchers that adherence may be diminished by the absence of clinician support [68], we did not find this to be the case in the context of this small trial. Objectively collected adherence rates are higher than those in other comparable eHealth interventions [70]. As we did not collect any parents’ feedback, we are uncertain whether their involvement optimized participant engagement and completion. We plan to explore this in an upcoming RCT. The study participants reported that they enjoyed biofeedback-based relaxation strategies. However, the potential additional benefit of combining biofeedback and CBT also remains unclear from these findings. Following more definitive evidence of its effectiveness, a head-to-head comparison of Starship Rescue and solely CBT and biofeedback interventions would be worthwhile to address this issue.

Despite being co-designed with its target audience, further minor modifications are warranted before the RCT to address some of the intervention’s technical aspects and improve its usability and acceptability. The fact that the mean length of time taken to complete the intervention exceeded expectations may be related to a number of issues. Extensive completion times for module 1 are likely to be because of participants being assisted by us to access module 1 during the onboarding process but not actually commencing or completing it until a later date. The delay in completing module 3 is likely to be related to participants getting stuck while playing the embedded platform game. Finally, although the requirement for real-world mastery of a chosen source of anxiety during module 4 may have proved challenging for users, this remains a vital means of generalizing therapeutic knowledge into practice [71]. Other key information from this pilot trial that will influence the subsequent RCT is the slow pace of recruitment from a single site by busy clinicians. For the RCT, recruitment from multiple sites is planned to ensure more timely data collection. The outcome measures used in this open trial appear suitable for use during RCT. Adherence to the intervention (defined by us as completion of all modules) was achieved by 67% (16/24) of the participants. This is comparable with previous studies of eHealth interventions, such as Smart, Positive, Active, Realistic, X-Factor Thoughts (60%), Cool Kids online (75%), and BRAVE online (85%), and a recent systematic review that identified a mean rate of completion of 64% for technologically delivered interventions for childhood anxiety and depression [42]. This is also encouraging, given the known association between module completion and outcomes for psychological eHealth interventions [72]. Although its reach may currently be limited by reliance on a physically worn heart rate monitor, emerging technology will likely permit biofeedback to be conducted via heart rate monitoring apps in the future.

The strengths of this study are the co-design of Starship Rescue with end users, the inclusion of participants with different long-term physical conditions, and the small amount of missing data. The limitations of the study include the small sample size; the study being conducted in a single location, which may affect the generalizability of results; and the use of only self-reported outcomes. Exploration of parent or clinician ratings would also be valuable for comparison with user-rated levels of anxiety, as would an exploration of the types of anxiety participants choose to address; more detailed exploration regarding the combination of biofeedback and CBT; and differences in completion and acceptability between users of different genders and ethnicities during an adequately powered RCT. Given the difference between intervention use between experimental and nonexperimental settings [73], investigation of the intervention’s use in a naturalistic setting would be useful. Future research would benefit from formal economic analysis to bridge the gap between researchers’ interests and policy makers [74].

Starship Rescue remains the only eHealth intervention specifically designed for treating anxiety and is evaluated with children and young people with long-term physical conditions. If future RCT results confirm the encouraging results from this pilot study, Starship Rescue has the potential to improve the short-term psychosocial well-being of this population by reducing psychological distress, improving quality of life, more optimal physical health management, reduced school absence, and improved social integration. In the longer term, it may also improve the rates of completed education, employment, and survival.