The recruitment strategies used in this study formed part of a larger study approved by the Swansea University School of Psychology Research Ethics Committee (1-2023-6798-6614). All participants provided written informed consent to take part. The data obtained were fully anonymized by the research team, and all participants were compensated with either Prolific funds or online shopping vouchers on completion of the larger study.

We developed the ACT Vet app using a co-design approach with veterans with lived experience of harmful gambling or PTSD and then adopted the Ionic Capacitor framework to achieve a research-viable product standard. ACT Vet is a self-guided app addressing the core components of ACT through seven modules: (1) an introduction that provides an overview of ACT Vet and ACT as a therapeutic practice, (2) Exploring thoughts around coping (cognitive defusion), (3) Control as the problem (acceptance), (4) Contact with the present (present moment awareness), (5) The self as the observer (self as context), (6) Living by your values (values), and (7) Committing to action (committed action ). The modules took approximately 15 minutes to complete, with a mandatory break period of 5 days taking place between completed and new modules. The ACT-specific module content took approximately 4 weeks to complete, and participants were encouraged to use the app for the remainder of the 10-week study.

The primary recruitment and retention strategies included online engagement, providing study details and links across social media sites: Facebook, X (formerly known as Twitter), and Instagram. Individual project pages across each site (Facebook, X, and Instagram) were created (in May 2024) with the purpose of information sharing using blog-type posts and project recruitment (Figure 1, left panel). Paid Facebook advertisements were used as they have been considered a useful tool in similar veteran-specific app-based recruitment efforts [32]. These advertisements were run for a total of 27 days. The textual keywords that the target audience needed to have on their profiles to be included are as follows: Armed Forces Day (United Kingdom), veterans’ benefits (military) or army men, school or university, British Army and employers, Royal Air Force Regiment, Royal Marines, veteran, His Majesty’s Armed Forces, and Royal Navy or British Army. Figure 1, middle panel shows the imagery and copy used for the advertisement.

Additionally, Prolific, which is an online recruitment platform where researchers advertise studies among a large pool of vetted participants, was used to screen potential participants who were UK military veterans aged ≥18 years. A dedicated project-specific website was also launched that housed information and links related to participating in the study.

Email addresses were collected from organization websites, and study details were sent. The research team carried out face-to-face meetings and visits with organizations such as the Forces in Mind Trust, National Health Service West Midlands Gambling Harms Clinic, the Royal British Legion, Combat Stress, Help for Heroes, Beacon Counselling Trust, GamFam, and GamCare. Physical and digital flyers (Figure 1, right panel) were created and circulated among organizations. Members of the research team also had personal contacts they used, including details of veterans who had previously agreed to participate in the study. Any veteran who agreed to take part was asked to encourage others who may have benefited from participating (snowball sampling).

The information regarding the recruitment source for each participant was documented and assigned a unique identifier, allowing the research team to track where participants had viewed and accessed the prescreening questionnaire (eg, Facebook, Prolific, or website). If eligible, previous military service numbers were also confirmed for the identification of spam or bots to ensure the integrity of the data. Unless provided by the recruitment platform, demographic information was collected at the baseline time point. Finally, it was advertised that, on completion of the 10-week program, each participant would be offered a high street voucher of £150 (US $189.77) for their time.

Our primary outcome measures included advertisement duration in days, cost in GBP (and the USD equivalent), online market reach or armed forces audience criteria, online post link clicks, and online post shares and reactions. Please note that online market reach differs depending on when digital marketing campaigns are conducted because of the underlying Facebook or Meta analytics. Means and SDs were reported for the continuous variable (respondents’ ages), and percentages and frequencies were reported for categorical variables.

Figure 2 shows the attrition from initial screening to week 10, and the completed time points are split by recruitment source. A total of 35 participants (n=31, 89% male; n=3, 9% female; and n=1, 3% undisclosed) completed in-app baseline time points, and 1 (3%) participant formally withdrew from the study. The mean age was 44.77 (SD 10.89) years, and all participants (35/35, 100%) were of White British ethnicity, with 49% (n=38) having a diagnosis of PTSD and none having a DSM-5 diagnosis of gambling disorder. A total of 24 participants (n=21, 88% male; n=2, 8% female; and n=1, 4% undisclosed) with a mean age of 45.29 (SD 10.70) years finished the week 10 time points and had a mean PTSD Checklist for the DSM-5 score of 40.92 (SD 20.12) and mean Problem Gambling Severity Index score of 5.47 (SD 3.59), indicating moderate risk of problem gambling. According to responses to the mHealth App Usability Questionnaire, participants provided high ratings of usability (mean 6.09, SD 0.76; range 4.18-7.00), indicating that the app was liked by veterans.

This paper outlines recruitment and retention strategies used for ACT Vet, a smartphone digital intervention for UK military veterans with PTSD and harmful gambling. Overall, Facebook (both unpaid and paid advertising) and Prolific were the most effective strategies for recruiting eligible participants. The combined alternative recruitment methods (hub visits, meetings, emails, and website) allowed for building interdisciplinary networks among researchers, service users, and practitioners but resulted in few participants. Challenges throughout recruitment remained and included low conversion rates from social media advertising, spam or bots, ineligible participants, participant attrition, and missing questionnaire time points. However, app usability ratings were high, and the app was well received.

Social media was effective in recruiting eligible participants, aligning with previous findings [30,32-34]. In this study, although we only ran brief advertisement campaigns, Facebook outperformed traditional strategies in the recruitment of eligible veterans. However, there was a noteworthy difference in link clicks through to eligible consent using paid Facebook advertising. Most of those who showed initial interest by clicking the advertisement link (n=1421) did not proceed through to eligible consent (n=1407, 99.01%). Even so, as evidenced by these results and highlighted by Aily et al [33], a low conversion rate from Facebook paid advertising can still contribute to a cost-effective recruitment strategy.

Prolific was also effective in recruiting and retaining eligible military veterans. Most of those who completed final questionnaire time points were recruited via Prolific (n=19, 79%) and helped the study access secondary categories of groups that may be hard to reach [39]. Prolific was also the most cost-effective compared to all other strategies. This is in line with previous work that argues that Prolific can be cost-effective and reliable for high-quality data [40,41]. Hence, our study shows that the combination of Facebook and Prolific is important in recruiting for clinical research, particularly military veterans with co-occurring PTSD and harmful gambling.

Notwithstanding the aforementioned advantages to the approaches used, we did encounter some recruitment and retention challenges. Few veterans were recruited through strategies such as face-to-face meetings, emails, or visits to organizations. Veterans can take up to 11 years to access support services and may face significant barriers and stigmatization when accessing treatment [1,3], which may have contributed to the limited interaction observed in this study. However, due to the anonymous nature of this study, it was difficult to measure whether more veterans who later accessed the online information were recruited because of our in-person visits. For this reason, it is advisable that future digital or online interventions maintain a physical presence among service users, charities, and hubs to build in-person networks and raise further awareness. Moreover, future recruitment initiatives should consider ensuring that long-term strategies and resources are dedicated to developing engaging content and other in-person activities with veterans.

In addition, the reliance on social media resulted in many spam responses and unreliable data, which may have contributed to the large attrition rate. Even among eligible participants, close to half (n=39, 49%) did not respond after requesting and being sent individualized app codes and did not go on to complete baseline measurements. These participants could not be contacted and, therefore, were considered to have withdrawn. Similar instances of attrition have been noted in UK-based military veteran digital interventions that have used social media for recruitment [19,20] and are considered a key challenge in clinical mHealth research. As recommended by Parkes et al [20], future work using social media for recruitment should also ensure that additional strategies and resources are made available to counteract any attrition from social media recruitment.

Data attrition inevitably resulted when participants did not complete required questionnaires when requested. This led to variation between program time and data points. In this study, there was an overall decrease in completion of all measures from baseline to each time point, yet the final week, week 10, had more completions than interim weeks such as weeks 4 and 6. One reason for this could be that participants stopped using ACT Vet after all steps had been completed within the app by the week 4 time point. While push notifications and email reminders were sent, this study still experienced a decrease in completed measures during those weeks. The opportunity to provide week 10 data was provided using an email link after a lengthy period despite also being the final time point, which may have resulted in an increased response rate. Thus, if the program length extends past the intervention length (ie, in study designs with extended follow-up), attrition is likely to occur without targeted initiatives to increase engagement.

A final retention challenge may have inadvertently stemmed from offering remuneration in a single large amount at the end of the study. That is, attrition may have been reduced had we adopted a progressive or escalating schedule of payments similar to those used in research on incentivized treatment attendance or adherence to interventions for related addictive behaviors [44]. This possibility warrants consideration in future mHealth intervention research.

There were some limitations to the recruitment and retention methods we used. First, participants recruited through Prolific were self-selected, which may increase the likelihood of bias and decrease the likelihood of obtaining a representative sample [35]. Despite this, data integrity measures were implemented throughout, such as requiring service numbers as evidence of veteran status and validating data quality. Second, as found in similar recruitment efforts, the use of online media platforms may have excluded veterans without social media or with limited experience using technology [32]. However, this study implemented mitigating strategies to reduce these issues [30]. Third, the range of colors, imagery, and designs used for the project-specific branding may have resembled army branding (ie, army green and camouflage patterns). There is a possibility that this branding impacted recruitment of veterans from other services such as the Royal Navy and the Royal Air Force, both of which have unique colors and branding. Finally, 89% (n=31) of all eligible recruited participants were male, which resulted in an underrepresentation of female veterans. Although female representation in the regular UK armed forces increased by 11.7% in 2024 [45], the veteran criteria for ACT Vet likely resulted in a sample that reflected representation in the UK armed forces, as previously noted by Williamson et al [32]. Furthermore, men are considered at high risk of harmful gambling [46], which may have contributed to the sex variance observed.

In conclusion, our findings showed the effectiveness of social media and online platforms compared to additional recruitment strategies in line with previous findings and recommendations for mHealth recruitment [30,33,34], particularly for military veterans [32]. Furthermore, these findings can be extended to populations with comorbid conditions, such as veterans with both PTSD and harmful gambling. In this paper, we offer some key recommendations for future recruitment efforts. First, face-to-face meetings and visits should be organized as early as possible to strengthen interdisciplinary relationships and networks, but they should not be solely relied upon for recruitment. Second, a strong online presence should be built across various platforms, and unpaid and paid strategies should be used. This will assist in project awareness and recruitment from populations that may be hard to reach, such as those with PTSD and harmful gambling. Finally, increased efforts should be made to diversify branding and advertising. For military veteran research, this should reflect all 3 UK regular armed forces services (Army, Royal Navy, and Royal Air Force), which, in turn, may also lead to an increase in participation from female veterans with comorbid disorders. It is hoped that these findings inform future recruitment initiatives in veteran and gambling research and mHealth research more broadly.