The app administered a brief EMA at random times during three prespecified periods (9 AM to noon, 1 PM-4 PM, and 5 PM-8 PM) via push notifications. The EMA comprised five core items to measure gambling episodes (since the last EMA), gambling cravings (current and since the last EMA), gambling self-efficacy (current confidence in ability to limit or stop gambling), and craving self-efficacy (current confidence in the ability to resist cravings) and up to an additional seven items depending on core item responses. These additional items were included to gain descriptive data about gambling cravings (eg, duration, frequency) that are largely absent in the corresponding empirical literature [43]. The EMA took approximately 1-2 minutes to complete, depending on the pattern of responses (the EMA items are given in Table S2 in Multimedia Appendix 1).

The app’s EMI feature involved an automatic recommendation to use any urge-curbing tip or activity (Figure S2 in Multimedia Appendix 1) at the end of any EMA in which participants reported that they had a current craving (ie, responded Yes to EMA item 1; Table S2 in Multimedia Appendix 1). The program was also available 24×7 for use on demand. Whenever participants used a tip or activity (regardless of whether use was recommended or on demand), they were asked to rate the intensity of their craving to gamble from 0 (mild) to 10 (severe) immediately before and after using the tip or activity.

Acceptability was assessed via a number of measures administered at postintervention, including (1) helpfulness of each urge-curbing tip or activity, and the relevance and burden of EMA items, via separate 11-point Visual Analogue Scales (VAS) from 0 (not helpful or relevant or burdensome) to 10 (very helpful or relevant or burdensome) [40]; (2) satisfaction with the intervention via the 3-item Client-Satisfaction Questionnaire-3 (CSQ-3; total score range 3-12) [46], with higher scores indicating greater overall satisfaction, participant needs met, and likelihood of future participant use; (3) impact of the intervention on participants’ awareness, knowledge, attitude, intention to change, help-seeking behavior, and behavior change in relation to gambling cravings via the 6-item App-Specific subscale of the Mobile App Rating Scale (MARS) [47], whereby the mean item score of 3 (range 1-5) indicates minimum acceptability [48]; and (4) a series of open-ended items assessing suggested improvements for any tip or activity that participants rated a 5 or less (out of 10) for helpfulness (How could [tip/activity] be improved?), any technical issues (Please comment on any technical issues experienced.), and general feedback about the app intervention (Do you have any other comments or feedback?) [40].

Feasibility was assessed via participant recruitment at baseline and retention at postintervention and follow-up evaluations. Feasibility was also assessed using several app use metrics, including EMA compliance and EMI compliance. EMA compliance was measured as the rate at which participants completed the EMA during the baseline and intervention periods. EMI compliance was measured as the rate at which participants completed any intervention content within 60 minutes of receiving an EMI recommendation to use the intervention because they reported a current craving to gamble during an EMA [38,49,50]. Feasibility was also assessed via intervention use more generally. Upon inspection of the data, any intervention use was stratified by EMA-prompted use (defined as intervention use within 60 min of completing an EMA, regardless of whether they were recommended an activity based on craving occurrence) and on-demand use (defined as any other intervention use). The term EMI use was employed for the subset of EMA-prompted use in which use occurred following an EMI recommendation to use an intervention based on craving occurrence (note that this is equivalent to the EMI compliance rate).

Preliminary effectiveness was assessed using changes in (1) real-time craving intensity measured by the aforementioned rating items administered immediately before and after using any urge-curbing tip or activity and (2) outcome measures completed at baseline, postintervention, and follow-up evaluations.

In the outcome evaluation, past-week gambling symptom severity (primary outcome) was measured using the 12-item Gambling-Symptom Assessment Scale (G-SAS) [51], with total scores (range 0-48) indicating mild (score of 8-20), moderate (score of 21-30), severe (score of 31-40), and extreme (score of 41-48) severity. In terms of secondary outcomes, gambling cravings were measured using the G-SAS Urge Subscale, which comprises the first 4 G-SAS items (score range 0-16) assessing the craving intensity, frequency, duration, and subjective control. Past-month gambling frequency (days) and expenditure (Aus $) on 6 gambling activities (electronic gaming machines or pokies, table games, racing, sports and event betting, number games, and informal private betting) were measured using a series of single items. Participants’ confidence in their ability to resist a craving to gamble (craving self-efficacy) and to limit or stop their gambling (gambling self-efficacy) were measured using separate 11-point VAS from 0 (not at all) to 10 (very confident) [43].

At baseline, participants reported their demographic information and whether they had a problem with 6 gambling activities (eg, table games and number games). Participants also completed the 9-item Problem Gambling Severity Index (PGSI) [52] to measure their past-year problem gambling status. Total PGSI scores (range 0-27) were used to indicate nonproblem (score of 0), low-risk (score of 1-2), moderate-risk (score of 3-7), and problem (score of 8 or higher) gambling.

The mean ratings for the perceived helpfulness of all urge-curbing tips and activities were higher than 5 out of 10, indicating above-average helpfulness. The highest rated overall were Delay and Distract (mean 7.41, SD 2.06), About My Urge (mean 6.73, SD 2.39), Change Your Thoughts (mean 6.55, SD 2.46), Get to Know Your Thoughts (mean 6.55, SD 2.72), and Urge Surfing (mean 6.55, SD 2.79). The mean ratings of the EMA’s perceived relevance were high (7.45 out of 10, SD 2.02) and burdensome nature were average (4.59 out of 10, SD 3.74).

On the CSQ-3, participants scored an average of 8.86 out of 12 (SD 2.05). Individual item mean scores of approximately 3 out of 4 indicated that participants were mostly satisfied with the intervention, that the intervention met most of their needs, and that they would likely use it again to manage their cravings.

Mean scores across the MARS app–specific subscale items (range 3.64-3.91 out of 5) met acceptability standards (score of 3). From highest to lowest rated impact, participants indicated that the intervention improved their awareness about the importance of addressing cravings, future help seeking for cravings, knowledge and understanding of cravings, attitudes toward addressing cravings, and intention to address and ability to manage cravings.

Overall, 16 participants rated the helpfulness of at least one tip or activity a 5 or below (out of 10) for managing cravings and were subsequently asked to suggest improvements on that tip or activity. Most of these participants tended to give a variation of “unsure” or no improvements required. Nevertheless, 1 participant suggested that Delay and Distract, Mindfulness, Urge Surfing, and Change Your Thoughts could be improved by providing more examples of “what has worked for others”. Furthermore, 1 participant suggested improving Talk to Someone by adding a live chat function to enable in-the-moment messaging support and 2 participants suggested improving the mindfulness- and relaxation-based activities by linking them to “a proper meditation/relaxation program” and incorporating additional audiovisual components.

The majority of participants (15/22, 68%) reported no technical issues with the intervention. Five participants reported that the audiovisual content was slow to load, 1 participant reported that the app exceeded their phone’s storage capacity, and 1 participant reported that the tips and activities incorrectly appeared as not completed after completion. Almost all of the participants (20/22, 91%) reported no EMA technical issues; however, 1 participant reported that they did not receive notifications and 1 participant reported that EMAs wrongly displayed as not completed after completion.

Half of the participants (11/22, 50%) provided general feedback about the intervention, with most of them (7/22, 32%) providing positive feedback. Specifically, 3 participants found the EMA notifications helpful daily reminders to stay on track and not gamble and 4 participants found the intervention content to be highly accessible and helpful, with 1 participant stating that the app was “a great way to have different tools available at any time” and another stating that “the app worked well, was easy to use, notifications were good, (and the) training things were helpful and informative.” Four participants provided negative feedback about the EMA notifications, whereby 2 participants requested tailoring capabilities to user-specified times (eg, “pay day”), 1 participant stated that the notifications “reminded me about gambling when I wasn’t thinking about it,” and 1 participant thought that the notifications were too frequent.

Over the 1-month recruitment period, 56 gamblers consented to participate in this study. Of these, 36 gamblers completed the baseline measures, 14 gamblers did not complete any baseline measures, and 6 gamblers completed only the baseline demographic measures. Of the 36 participants who completed the baseline measures, 9 participants did not download the app, 1 participant downloaded the app but did not use it, and 4 participants did not use the app beyond the first week of the study and did not complete the postintervention evaluation. The remaining 22 participants used the app and completed the postintervention evaluation, with 21 of these participants completing the 1-month follow-up evaluation. Of the 36 participants who completed baseline measures, 61% (22/36) completed postintervention evaluations and 58% (21/36) completed follow-up evaluations. There were no significant differences in any baseline variable between participants who did or did not complete the postintervention evaluation.

The EMAs had a compliance rate of 68% in the baseline period (mean 14.27, SD 5.68 out of 21 EMAs), 47% in the intervention period (mean 54.05, SD 33.37 out of 84 EMAs), and 51% overall (mean 54.05, SD 33.37 out of 105 EMAs).

The EMI compliance rate was 15%, as participants used an intervention 13 out of the 87 times that they were recommended to do so based on craving occurrence.

Overall, 19 participants used the intervention at least once during the intervention period. Table 3 presents the descriptive statistics for intervention use. The intervention was used a total of 166 times (median 7; range 1-33), including 59 EMA-prompted uses (defined as intervention use within 60 min of completing an EMA), of which 13 were EMI uses (defined as intervention use within 60 min of completing an EMA and following an EMI recommendation to use an intervention because the participant reported a current craving). Of the 59 EMA-prompted uses, participants used the intervention once between EMAs on 29 occasions, 2 times on 10 occasions, 3 times on 2 occasions, and 4 times on 1 occasion. In contrast, there were 107 on-demand uses (defined as any other intervention use). The most used intervention content included About My Urge (22 uses), Talk to Someone (19 uses), and Mindfulness (18 uses).

On the basis of the 22 participants who completed the postintervention evaluation, Table 4 presents descriptive statistics for EMA outcome variables. Compared with the 1-week baseline period, there was a 71% reduction in the average number of gambling episodes and a 72% reduction in the average number of craving occurrences, as reported in the 4-week intervention period. Furthermore, craving self-efficacy and gambling self-efficacy increased from mean ratings of 5.85 (SD 2.95) and 4.82 (SD 3.01) out of 10 (where 10=complete confidence in ability to resist a craving and to limit or stop gambling) over the baseline period to mean ratings of 6.63 (SD 2.61) and 5.72 (SD 3.02) over the intervention period, respectively.

On the basis of the 22 participants who completed the postintervention evaluation, the results of clustered paired-sample t tests revealed a significant decrease of 5.4% in momentary craving intensity from immediately before to immediately after any intervention use (P=.01), with a medium effect (Cohen d_z=−0.64; 95% CI −1.13 to −0.14). The results demonstrated a 7.5% decrease in craving intensity for EMA-prompted use (P=.03; Cohen d_z=−0.72; 95% CI −1.34 to −0.06), and more specifically, a 10.5% decrease for EMI use (P=.01; Cohen d_z=−1.29; 95% CI −2.48 to −0.03). In contrast, there was a 4.5% decrease for on-demand use (P=.01; Cohen d_z=−0.66; 95% CI −1.16 to −0.14). There was no added benefit of using multiple interventions between EMAs, as the results demonstrated an 8.6% decrease in craving intensity after one use (P=.02; Cohen d_z=−0.94; 95% CI −1.67 to −0.17) compared with a 7.7% decrease after more than one use (P=.29; Cohen d_z=−0.63; 95% CI −1.49 to 0.28).

Descriptive statistics for gambling symptom severity, cravings, frequency, expenditure, and self-efficacy measured at baseline (n=36), postintervention (n=22), and follow-up (n=21) are presented in Table 5. At the group level, participants showed significant reductions in mean G-SAS gambling symptom severity (d=0.61 and 0.75; P=.01 and .003), G-SAS cravings (d=0.49 and 0.55; P=.03 and .02), gambling frequency (P=.01 and .004), and gambling expenditure (P=.04 and .003) from baseline to postintervention and follow-up, respectively. In addition, participants showed a significant increase in mean craving self-efficacy and gambling self-efficacy from baseline to postintervention (d=0.66 and 0.60; P=.01 and .01) and gambling self-efficacy (d=0.49; P=.04) but not craving self-efficacy (P=.20) from baseline to follow-up.

Table 6 presents clinically significant change on these measures from baseline to postintervention and follow-up. At an individual level, approximately a quarter of participants were recovered at postintervention and follow-up, respectively, on G-SAS gambling symptom severity (6/22, 27%; 6/21, 29%) and G-SAS gambling cravings (6/22, 27%; 7/21, 33%), the majority were unchanged (symptom severity: 13/22, 59%; 11/21, 52%; cravings: 14/22, 64%; 14/21, 67%), and a small number of participants were deteriorated at postintervention (symptom severity: 1/22, 5%; cravings: 2/21, 9%); however, none were deteriorated at follow-up. The majority of participants were unchanged at postintervention and follow-up on gambling frequency (20/22, 91%; 18/21, 90%), gambling expenditure (19/22, 86%; 17/21, 81%), craving self-efficacy (18/22, 82%; 17/21, 81%), and gambling self-efficacy (18/22, 82%; 18/21, 86%); however, none were deteriorated on these measures.

Overall, the app intervention was rated favorably across quantitative and qualitative measures of acceptability. Participants indicated that they were mostly satisfied with the intervention, as it met most of their needs and could be used to manage their cravings in the future, based on mean CSQ-3 scores. Furthermore, participants indicated that the intervention improved their awareness, knowledge, attitude, and intentions to address cravings and, importantly, their ability to manage cravings, based on mean MARS item scores. Qualitatively, the majority of participants provided positive feedback about the intervention, particularly in relation to its accessibility, helpfulness, and effectiveness in preventing gambling episodes.

Promisingly, participants rated all of the urge-curbing tips and activities as above average for helpfulness in managing cravings. Consistent with the evidence base for gambling treatments, in which CBT techniques are considered best practice [7,41,61,62], the highest rated content involved CBT techniques, including psychoeducation (About My Urge); cognitive reappraisal (Get to Know Your Thoughts and Change Your Thoughts); and relapse prevention techniques, including distraction (Delay and Distract) and a mindfulness practice to surf the wave of a craving without acting on it (Urge Surfing). Participants suggested that this content could be further improved by providing more concrete examples of skill implementation. The lowest rated content included relaxation-based or imaginal techniques (Progressive Muscle Relaxation, Belly Breathing, and Brief Imagery), with 2 participants suggesting their improvement via linkage with related programs outside of the app. Taken together, these findings suggest that intervention content variability and examples of skill implementation are important for people to find, and learn to apply, what works best for them.

With respect to the EMA, participants considered them highly relevant but somewhat burdensome, which represents an improvement from the initial usability testing [40] in which they were rated in the average range on both domains. To reduce its level of burden, the frequency of EMAs could be reduced or, as 2 participants suggested, made personally customizable to specific times (eg, pay day). Indeed, the combined results of usability testing and this trial indicated that tailoring capabilities were considered particularly important for consumers to enable personalized check-ins and treatment.

Recruitment and retention have posed a challenge in studies examining app-delivered interventions for mental health [61], particularly for problem gambling [29]. Despite extensive recruitment efforts, only 36 people completed baseline measures, with approximately 60% of these people completing postintervention (n=22) and follow-up (n=21) evaluations. Retention rates in this trial were lower than those of other feasibility trials of app-delivered EMIs for alcohol and substance use, where rates range from 63% to 90% [35,38]. Although not directly comparable, retention was on par with the only other feasibility trial of a gambling intervention app (60%) [26]. Future research may benefit from examining consumer profiles to enable targeted recruitment campaigns and tailored interventions to particular demographic groups, the latter of which may also facilitate trial retention through improved engagement with the app [62].

In this trial, the feasibility was measured via compliance with the EMA and EMI features. The overall compliance was 51% for EMAs and 15% for EMIs. Although there are no directly comparable gambling app intervention studies, previous gambling EMA studies have reported variable compliance rates from 50% to over 90% [43,63]; however, notably, EMA completion was financially incentivized in these studies. As financial incentives are not feasible for ongoing app interventions in real-world contexts, it would be useful to examine ways to improve EMA compliance. In addition, EMA compliance may have been reduced in this study as participants could access intervention content on demand without needing to complete a more time-consuming EMA. While this is a limitation of providing a hybrid push or pull intervention, this approach supports consumers to access interventions regardless of their awareness of vulnerable states or motivation to seek support [45]. Furthermore, EMA compliance may improve outside of a research context, as it is envisioned that the EMA feature of future iterations of the intervention for real-world use would be briefer (eg, to only measure current craving occurrence) and customizable regarding the frequency and timing of administration (eg, to user-specified times). The EMI compliance rate in this study was also low, however, as participants used the intervention only 13 of the 87 times that they were recommended to do so. Low EMI compliance may reflect the limitations of the app platform, whereby intervention content cannot be automatically delivered (rather than just recommended). Although this limitation enabled a more nuanced evaluation of the intervention’s preliminary effectiveness under different conditions of use, future app iterations may benefit from using platforms that can automatically deliver an EMI at times of need.

Across the 4-week intervention period, participants used the intervention a total of 166 times (median 7 uses; range 1-33), including 59 EMA-prompted uses and 107 on-demand uses. This rate of intervention uptake (7 uses over 4 weeks) appears comparable with that of traditional face-to-face gambling services, in which treatment typically involves 1 session per week for 10 weeks [64]. Despite similar uptake rates, we might expect higher app use given its 24×7 availability and evidence of EMI underutilization. Although apps can overcome a number of barriers to accessing face-to-face services (eg, geographic constraints) [11,12], future research should explore app-specific barriers to use at times of identified need. The higher rate of on demand (pull) intervention use compared with EMA-prompted and EMI (push) use suggests either that the EMA item measuring craving occurrence was not adequately identifying the times of vulnerability and/or that participants knew when they needed to access an intervention and were motivated to do so [45]. To address the latter possibility, future app interventions may benefit from incorporating additional user-initiated EMAs [18,34], whereby participants can record a gambling craving or episode at any time in the app.

A unique strength of EMA and EMI data is the capability to explore real-time intervention effects in people’s everyday lives. Promisingly, descriptive EMA data showed that, compared with the baseline period, there was an approximately 71% to 72% reduction in the average number of gambling episodes and craving occurrences and an approximately 1-point increase in the mean craving self-efficacy and gambling self-efficacy on an 11-point VAS, reported in the intervention period. In addition, the results showed that the intervention led to a medium decrease of 5.4% in real-time craving intensity from immediately before to immediately after use. This effect increased with a large decrease of 10.5% for EMI use (ie, where participants had a current craving to gamble) and 7.5% for EMA-prompted use (ie, use within 60 min of an EMA regardless of whether participants had a craving) and reduced with a medium decrease of 4.5% for on-demand use (ie, any other intervention use), which intuitively supports the intervention’s increased effectiveness in moments of vulnerability. The results also suggested an optimal dosage of 1 intervention between EMAs, as there was no added benefit of multiple uses.

Given the small sample size and limited power to detect significant effects, this study also explored the intervention’s preliminary effectiveness at both the group and individual level. At the group level, participants displayed significant medium-to-large reductions in mean gambling symptom severity and cravings and reductions in gambling frequency and expenditure at postintervention and 1-month follow-up evaluations. Participants also displayed significant medium-to-large increases in gambling self-efficacy at postintervention and follow-up and craving self-efficacy at postintervention. At an individual level, more than a quarter of participants were considered recovered or improved on G-SAS gambling symptom severity (8/22, 36%; 10/21, 48%) and G-SAS gambling cravings (6/22, 27%; 7/21, 33%), at postintervention and follow-up respectively, with the majority remaining unchanged, which compares with more than a half of participants (n=55, 64%) on gambling symptom severity in a randomized trial of its parent program (GamblingLess) [41]. A small proportion of participants in this study were also recovered or improved on gambling frequency, expenditure, gambling self-efficacy, and craving self-efficacy (2/22, 9% to 4/22, 19%). By the end of this study, only 1 participant was considered deteriorated on craving self-efficacy but no other measure. Although it is possible that these results reflect a highly motivated sample, natural recovery, or recovery resulting from concurrent treatment, these findings provide preliminary evidence for the effectiveness of the intervention, which would likely increase further when combined with a fuller suite of gambling interventions.