Those who seek self-directed treatment for many mental health problems are spoiled for choice. In fact, there is so much choice for some disorders that treatment selection itself can be overwhelming [1]. Paper-and-pencil workbooks, web programs, and mobile apps permeate the self-help space, particularly for mood and anxiety disorders [2-4]. Generic mental health apps, including Headspace (Headspace Inc) and Calm (Calm.com, Inc), are also highly popular [5,6]. People with gambling problems are not nearly as well-supported in their specific needs, some of which are predominant in gambling, such as financial concerns. This gap is unfortunate, given that there is a high demand for self-directed gambling treatment. Most gamblers who attempt to change do so independently [7], and many prefer to use self-directed interventions for the privacy, ease of access, and ability to control their pace and level of engagement [8-11]. The current paper describes and contextualizes the development of a self-directed, tailored mobile app for gambling problems specifically designed to enhance user engagement.

Evidence-based self-directed treatment for gambling problems first took hold in the 1990s and 2000s, mainly in the form of cognitive behavioral bibliotherapy [12]. In the 2010s, much of the same content was delivered via the web to expand accessibility and cost-effectiveness [13]. Multiple theories and treatment approaches were implemented in these resources, many of which were offshoots of cognitive behavioral therapy (CBT) or third-wave modalities. Newer technologies, like mobile apps, virtual reality, and artificial intelligence (AI), have become a central focus of self-directed interventions for gambling problems in the 2020s [14]. These emerging modes of delivery are especially promising because, in addition to offering all the benefits of 2010s-style web-based interventions, they increasingly permit tailored, responsive, and interactive content rather than static, generalized, or read-only resources. As it stands, however, very few evidence-based options are publicly available to treatment-seeking gamblers.

In a review, Merkouris [15] identified 6 evidence-based mobile apps for gambling problems. Most were based on preexisting paperback or web-based workbooks (eg, study by Dowling et al [16]). However, only 1 French-language app was publicly available. The most thoroughly researched app, GamblingLess: Curb Your Urge (Deakin University), specifically targeted transient gambling urges with an array of cognitive, behavioral, and mindfulness-based strategies [17,18]. Preliminary results indicated good acceptability and effectiveness, albeit low uptake and compliance. The same research group has published protocol papers for the development of similar mobile apps that vary in their intended audience [19-21]. Other mobile apps for gambling problems are publicly available, but most are solely abstinence-based and do not rely on a recognizable treatment model [22]. For example, most available apps consist only of an abstinence time tracker, while some also include reminder notifications and motivational quotes. Some newer apps contain gambling website blocking software, gambling-related information, and directories of external support services, but very few were developed in conjunction with people with lived experience [23]. Taken together, there are not many viable mobile apps that target gambling problems, but preliminary research on a select few is encouraging.

Perhaps the greatest challenge in offering self-directed interventions for mental disorders, including gambling problems, has been low user engagement [24]. Engagement in this context is often measured in terms of objective and subjective indicators of treatment use, such as the number of activities completed or ratings of subjective appeal [11,25]. Engagement is not simply adherence to an intervention, although adherence is often considered one piece of engagement [26,27]. Rather, engagement likely comprises a combination of behavioral, cognitive, and affective factors, including enjoyment, motivation, and perceived relevance [27,28]. Perski and colleagues [11] proposed that subjective and objective engagement is moderated by the content and delivery of the intervention, the real-world context of its use, and the behavior it targets. Treatment outcomes are, in turn, moderated by user engagement. The recent development of tailored and adaptive interventions has been, in large part, a response to the problem of low engagement by improving intervention content and delivery.

The problem of low user engagement has been underscored in the gambling field, as intervention usage data has revealed just how modest user engagement tends to be. In our recent trial of a static self-directed web program for problem gambling [29], we found that 41% (128/313) of participants did not complete any modules, but those who completed more modules experienced greater reductions in gambling problems. Similar findings were documented in our own past work [30] and that of others [15]. This general observation is somewhat surprising given how positively users rate such interventions, and how self-directed change directly addresses barriers to traditional face-to-face treatment, such as stigma and lack of privacy [9,10].

The primary aim of this study was to contribute to the growing literature on evidence-based mobile apps for gambling problems by developing a personalized self-directed mobile app intervention grounded in content from our previously developed static workbook. The intervention was designed to facilitate greater engagement relative to static self-directed interventions by making the content more appealing and the delivery more responsive to individual needs. We aimed to conduct initial feasibility and usability testing with a small sample of Canadian adults with past or present gambling problems and solicit their feedback to inform subsequent improvements and testing. Given the formative nature of this research, effectiveness and longer-term engagement were not assessed at this stage.

This study received ethical approval from the Conjoint Faculties Research Ethics Board at the University of Calgary (REB23-0794). This approval authorized participant recruitment, informed consent, collection and storage of app usage and survey data, and participation in a focus group on Microsoft Teams. Participants were remunerated with up to CAD $40 (US $29.46) in electronic gift cards for their participation. Deidentified data from Qualtrics, MetricWire (MetricWire Inc), and Microsoft Teams were stored separately from personally identifying information in an encrypted file.

Canadian adults with past or present gambling problems were recruited over 3 months via email contact from researchers. These email invitations were sent to 100 participants from a previous gambling self-help trial [29] who had volunteered to be contacted for future research. We received 71 responses to an eligibility survey on Qualtrics, of which 33 were eligible, consented, and enrolled. Participants were eligible if they resided in Canada, were aged 18 years or older, and scored 5 or greater [31] on the Problem Gambling Severity Index (PGSI) [32]. Those with past problems were asked to respond to the PGSI based on their gambling at the peak of their problem instead of the typical past-year time frame. This approach was used to ensure participants without current gambling problems previously had a problem severity consistent with the app’s likely audience, that is, people with moderate to severe gambling problems [33].

Most of those excluded from this study were excluded because they did not complete the eligibility survey in full or did not respond to the initial email contact after they were found to be eligible. Furthermore, 3 participants were subsequently dropped because they did not create an account with the mobile app and did not respond to follow-up emails.

Enrolled participants were emailed a link and brief instructions to download and register with the third-party app, labeled Catalyst by MetricWire, in the Apple App Store and Google Play Store. Users interacted with our intervention via the Catalyst mobile app on which the intervention was mounted. We selected this platform to host our intervention because it was used successfully by our colleagues to develop a similar gambling intervention in Australia [16,18].

Participants were not given specific instructions for engaging with the app content but were informed that it was still under development, and the researchers would be interested in their feedback. Participants had access to the app for 2 weeks, after which they completed a brief feedback survey on Qualtrics and were compensated with an electronic gift card valued at CAD $20 (US $14.73). Then, 8 participants were randomly selected via a random number generator and invited to participate in a focus group facilitated remotely by author BWB and 2 research assistants on Microsoft Teams. Focus group discussions used a semistructured interview guide (Multimedia Appendix 1) to elaborate on user experiences with the app as well as their quantitative feedback. Because of ethical constraints, participants’ survey responses and focus group contributions were not linked to one another. Participants who provided written or oral feedback were compensated with an additional electronic gift card valued at CAD $20 (US $14.73).

No generative AI tools were used in any portion of this manuscript, including the generation, analysis, or presentation of language, content, data, or references. Quantitative data were analyzed using R and RStudio [50]. Results include means and SDs for continuous data as well as counts and percentages for categorical data. There were no missing data on the PGSI or uMARS. App usage data were incomplete by design, as not all content was necessarily used by all participants, and some usage (eg, viewing read-only resources) could not be captured by the platform. Accordingly, analyses of app engagement were conducted using available-case denominators specific to each feature, and no imputation was performed.

Given the formative nature of this study, feasibility and acceptability were primarily evaluated with the uMARS. Consistent with previous use of the uMARS in early-stage app evaluations [49], mean subscale scores of 3 or greater out of 5 were treated as an acceptability heuristic rather than a hypothesis test. Engagement and adherence metrics were not prespecified as stop-go feasibility thresholds; instead, these metrics were summarized descriptively to inform targets for subsequent iterations. The primary metrics of interest were diary completion rate, activity completion rate, and median activity duration.

Qualitative survey data were subjected to a simple content analysis. Responses were reviewed line-by-line, and codes were inductively generated. Codes were then grouped into categories of feedback based on various app features, and minority views were retained. Coding was done by author BWB in consultation with the research assistants who cofacilitated the focus group. Survey data were analyzed separately from the focus group because we did not have ethical approval to link specific survey responses to focus group participants. The intention of these analyses was to contextualize or provide brief examples of quantitative data rather than to add significant new information.

The sample (n=30) had a mean age of 44.0 (SD 12.4) years. Participants were mainly from Ontario (n=10, 33%) or Alberta (n=7, 23%), but all 10 Canadian provinces except Prince Edward Island were represented by at least 1 participant. The majority were White (n=22, 73%), employed (n=19, 63%), female (n=18, 60%), and in a married or common-law relationship (n=16, 53%).

In terms of gambling behavior, 26 (87%) participants regularly gambled in public settings and 12 (40%) in private settings. Additionally, 24 (80%) participants regularly gambled online. They engaged in about 4.6 (SD 2.5) types of gambling, the most popular of which were instant tickets (n=23, 77% participants), lottery tickets (n=21, 70%), video lottery terminals or slot machines (n=19, 63%), bingo (n=18, 60%), and raffles (n=16, 53%). Most participants (n=22, 73%) had ongoing (vs historical) gambling problems. The mean PGSI score of the full sample was 12.7 (SD 6.1), and there was no difference in terms of problem status (P=.55). Most participants (n=24, 80%) had previously received some form of support for gambling problems outside the context of a research study, but all had participated in the self-guided web intervention trial from which they were recruited. Only 5 (17%) participants reported they had used a mobile app similar to the current one, and 7 others heard of similar apps but never used them.

Feedback was gathered from participants in 3 ways—general feedback on Qualtrics (n=30, 100% response rate), specific feature ratings within the app (n=15, 50% response rate), and a 1-hour virtual focus group. Regarding the focus group, 10 participants were randomly selected and invited to participate. Of those 10, 8 consented to the focus group, 6 ultimately attended, and 2 provided written feedback via email instead. Furthermore, 2 participants did not respond to the invitation.

Quantitative feedback on Qualtrics was primarily collected via the uMARS (Table 2). Of the 4 objective domains, functionality was rated the highest on a 5-point scale, followed by information, aesthetics, and engagement. The highest-rated subdomains were information credibility, ease of use, and gesture functionality. The lowest-rated subdomains were customization, interactivity, and entertainment value. All subdomains except customization surpassed the acceptability heuristic threshold of 3.0.

Overall, the mean star rating for the app was 3.6 (SD 0.8) stars out of 5. Out of 30 users, 7 (23%) indicated they would be somewhat likely to pay for the app. The remainder were either ambivalent (9/30, 30%), somewhat unlikely (5/30, 17%), or extremely unlikely (9/30, 30%); none were extremely likely. In terms of projected use over the next year, if access were available, 3% (1/30) said never, 17% (5/30) said 1 to 2 times, 40% (12/30) said 3 to 10 times, 33% (10/30) said 10 to 50 times, and 7% (2/30) said more than 50 times. The dispersion of responses was similar when users were asked if they would recommend the app to others with gambling problems; 3% (1/30) said they would not recommend it to anyone, 13% (4/30) said they would recommend it to very few people, 43% (13/30) said some people, 10% (3/30) said many people, and 30% (9/30) said they would definitely recommend it to anyone with gambling problems.

As part of the uMARS, participants were then asked to rate the perceived impact the app would have in various domains. Those who endorsed a clear or strong positive impact (vs somewhat, minimal, or no positive impact) were 63% (19/30) for knowledge about gambling problems, 43% (14/30) for further help-seeking, 40% (12/30) for attitude toward change, 40% (12/30) for motivation to change, 33% (10/30) for importance of change, and 30% (9/30) for concrete change in gambling behaviors. Across all domains, 1 person consistently reported no anticipated positive impact, but the remainder endorsed minimal or somewhat positive impacts. The user with uniformly negative ratings did not provide any qualitative responses to elaborate on their ratings.

In addition to the uMARS, participants were asked to select up to 5 of the most and least important aspects of the app for achieving the primary goal of reducing gambling problems (Table 3). The response options were derived from the uMARS. The aspects of greatest importance were ease of use, interactivity, and quality of information. The aspects of least importance were entertainment value, invitation of user feedback, and overall layout. Aspects of greater disagreement included entertainment value, visual appeal, customization, and amount of information, as these were all rated among the most and least important by at least one-third of users. Similar response variability was observed in the open-ended feedback, which was offered by 23 of 30 (77%) respondents (Table 4).

Feedback was elicited from 15 of 30 participants (50%) within the app on the final day of enrollment. Ratings are summarized below (Figure 5) in terms of the percentage of users who identified each activity as somewhat or very helpful (vs minimally or not at all helpful). All interventions and resources were completed by at least 12 of the 15 users who provided in-app feedback (80%).

The sample was too small to evaluate whether focus group attendees and nonattendees differed in their app ratings, so we could not determine whether focus group responses were biased (for or against) the app. However, attendees provided both positive and constructive feedback that was largely consistent with open-ended survey responses.

By far, the most salient feedback from the focus group was the need to expand opportunities for user choice. This conclusion was largely driven by the group’s high variability in individual preferences, which mirrored the variability observed in the open-ended survey responses as well as the lower customization ratings on the uMARS. For instance, some strongly preferred thought-provoking and open-ended reflections, while others gravitated toward more efficient checkboxes or multiple choice. The same pattern was observed with respect to information medium, such that some preferred audiovisual resources, whereas others preferred text. Similarly, some liked the consistency of check-in questions to reduce cognitive load, while others felt disengaged by the repetition. Many users found the check-ins to be the most helpful part of the app, particularly to track spending and remain focused on treatment goals, but 2 strongly preferred other activities and resources. Some users felt overwhelmed by the amount of information requested or provided by the app, while others desired more data tracking or informative resources. One user provided suggestions for modifications to existing resources (eg, display all the same information in both text and visual format). Moreover, 2 users would have liked more time with the app to track their gambling behaviors, and 1 suggested gamification to enhance engagement. Participants provided some feasible technical suggestions for implementing the choice (eg, constantly available rather than time-limited expenditure log).

Areas of greater agreement included the frequency of notifications, utility of tracking and graphing expenditures, and rating scales for mood and urges. Most users found the frequency of notifications (ie, once or twice daily) to be just right. A number of minor technical suggestions were also provided to further enhance ease of use (eg, including an orientation page for first-time users and editing how buttons link to various pages within the app).