Mindfulness involves (1) attentional monitoring of present-moment experience (eg, thoughts, feelings, and sensations) and (2) orientation toward this experience with acceptance and nonjudgment [1]. Mindfulness-based interventions (MBIs) aim to train these skills and have been shown to improve a range of psychological and physical health outcomes in both clinical and nonclinical populations. For example, there is evidence from meta-analyses of randomized controlled trials suggesting that MBIs can reduce depression and anxiety or stress in young people [2], lower pain intensity in patients with chronic pain [3], and reduce symptoms of posttraumatic stress in people with and without a diagnosis [4].

Despite such efficacy, there are numerous challenges in accessing and delivering MBIs, including geographical, logistical, and financial constraints as well as a lack of trained mindfulness teachers [5,6]. For example, MBIs are typically face-to-face, multisession, and facilitated by expert interventionists, such as the mindfulness-based cognitive therapy (MBCT) course that is traditionally delivered by dedicated instructors in 8 weekly 2-hour group training sessions [7]. The translation of MBIs into digital formats has the potential to overcome these constraints, and it is encouraging that early evaluations of digital MBIs report beneficial effects that are comparable with those found in traditional in-person programs [8,9].

However, unfortunately, the success of digital MBIs is reliant on adequate engagement, which remains a crucial challenge. Engagement refers to the investment of energy in an activity and includes physical (ie, actual performance, which researchers often rely on when examining engagement using objective behavioral metrics [10]), affective (ie, affective reactions), and cognitive (ie, selective attention) elements [11]. For example, reviews of digital MBIs have found that between 8% and 52% to 60% of participants do not complete all sessions [9,12]. Although low engagement is a common issue in digital mental health interventions generally [13]—for example, the pooled completion rate from studies of apps for depressive symptoms is 52% [14]—it is particularly important in mindfulness training as regular practice is essential to develop mindfulness skills. Time spent practicing mindfulness at home is related to increases in levels of mindfulness and, in turn, improvements in psychological functioning [15]. Similarly, those who report high levels of engagement with digital MBIs report greater improvement in outcomes than those who do not [12].

Given that the success of digital MBIs is related to engagement and engagement tends to be low with digital MBIs, understanding the factors that facilitate or act as barriers to engagement with these interventions is crucial to promote engagement and opportunities to benefit. Past research has suggested that there is a range of factors that influence adherence to digital MBIs [5], including accessibility (eg, across devices and populations with different needs), tailoring (eg, of content to individual needs), and difficulty (eg, sustaining attention). In one study, after engaging with a digital MBI, students with no meditation experience reported that the top 3 obstacles to practice from a checklist of common challenges were meditation feeling like “just another task,” “feeling distracted,” and “feeling sleepy” [16]. However, the use of closed-response questions in such research potentially prohibits the development of a detailed understanding that is grounded in people’s own perspectives regarding aspects that help them engage and hinder them from engaging [17].

A more detailed approach using inductive qualitative analysis examined factors that hindered or facilitated the engagement of 16 health care professionals who participated in a self-help MBI (participants could choose a printed book or a web-based program) [18]. The results indicated that longer practices, arising negative thoughts, and self-criticism were key hindrances, and shorter practices, motivation to reduce stress, and feelings of control over thoughts were key facilitators. However, over half of the participants opted for the book-based intervention in this study, and themes identified from engaging with the web-based and book-based MBIs were combined. Although the authors reported that themes were comparable across intervention types, it is possible that barriers and facilitators specific to the web-based version were obscured by those common to both. Therefore, it is unclear whether these themes would apply to typical digital MBIs as well as to other populations (eg, groups who are vulnerable to or experiencing clinical-level concerns or for whom initial engagement is lower).

Although some studies have reported on factors that can influence engagement with digital MBIs, they rarely build a deep understanding of users’ experiences or do so systematically. User-centered design approaches (such as the person-based approach [19]) emphasize that understanding how people use digital MBIs and identifying core barriers to and facilitators of engagement are important first steps in intervention development, which suggest key design objectives to ensure interventions are relevant, acceptable, and engaging to target users before significant investment is made in evaluation and implementation [20]. This is particularly important in the context of digital mindfulness interventions as, unlike most digital health interventions, engagement with the digital content is designed to facilitate completion of a concurrent nondigital target behavior that is metacognitive in nature (eg, an experiential mindfulness exercise) [11]. As factors influencing engagement vary across different target behaviors, clear guidance is needed to understand which are directly relevant to and most prominent in digital MBIs specifically.

This review aimed to synthesize qualitative evidence on individuals’ experiences of using digital MBIs targeting psychosocial variables (eg, anxiety, depression, distress, and well-being) to identify key barriers to and facilitators of engagement. We chose to perform a rapid scoping review of qualitative data as (1) factors influencing the effects of interventions are often rooted in variations in attitudes, opinions, thoughts, feelings, and behaviors and, therefore, best explored through qualitative study [21]; (2) qualitative evidence is necessary to understand engagement in its entirety (ie, its physical, cognitive, and affective components [11]); and (3) it ensures that existing evidence can be incorporated into the early stages of intervention development and implementation [22,23]. The knowledge generated from this review will inform the evaluation and development of new and existing digital MBIs, helping them overcome some of the challenges that individuals face when engaging with these interventions.

We adhere to the Enhancing Transparency in Reporting the Synthesis of Qualitative Research guidelines [24] in reporting this review, and the review itself was guided by the Cochrane Rapid Reviews Methods recommendations [25] and PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews; Multimedia Appendix 1 [26]). We developed and preregistered an a priori protocol that specified the review questions (What are the key barriers to and facilitators of engagement with digital MBIs targeting psychosocial variables? How have interventions addressed and used these barriers and facilitators in the past, and in what ways could interventions address and use them in future?); participants, intervention, comparison, outcome, and study design; electronic database; search strategy; inclusion and exclusion criteria; and data charting form [27].

The inclusion and exclusion criteria were developed to identify qualitative explorations of individuals’ perspectives and experiences of using digital MBIs designed to improve psychosocial variables (Textbox 1). We excluded studies that did not refer to a digital web-based intervention (eg, a biofeedback headband and device based on vapor, light, and sound, both designed to support mindful breathing) and studies of interventions in which mindfulness was not the main component (eg, an intervention composed of 3 evidence-based techniques: cognitive behavioral coaching, motivational interviewing, and mindfulness). We specified that sample age ranges must at least partially overlap with 16 to 35 years as this is the target age group for our own intervention development. We defined digital MBIs as those delivered via the web by the technology itself (eg, hardware and electronic devices, software, and websites) rather than by health care professionals remotely [28]. Human support (eg, answering questions; providing feedback; and offering coaching, orientation, or check-in sessions) was permitted where the support was considered supplementary to the delivery of content, and we reported on the presence and format of such support in each included study. We focused on peer-reviewed papers as they would have received some initial quality assessment. Nonreporting bias [29] was minimized in this review as its focus was on generating themes related to engagement rather than estimating effects (ie, we did not extract quantitative results and included studies with no reported quantitative outcomes).

In consultation with an information specialist (psychology librarian who has extensive training in implementing structured database searches), we developed a comprehensive search strategy to identify articles with at least one keyword related to mindfulness, digital, user experience, and psychosocial variables in its title or abstract (Textbox 2). Keywords for psychosocial variables were derived from models of disordered eating [30] (ie, specific focus for our own intervention development), with added terms to broaden the search for all psychosocial variables (eg, affect, mood, distress, and well-being).

We uploaded the search results to Covidence (Veritas Health Innovation), a web-based systematic review software, to streamline the screening process. Consistent with guidance from the Agency for Healthcare Research and Quality [31], we started with a pilot phase to calibrate and test the eligibility criteria. In total, 2 researchers independently screened a random selection of 50 studies (10% of the records) and then met to resolve discrepancies (Multimedia Appendix 2 [32]). The first author screened the remaining titles and abstracts. All potentially eligible records were obtained as full-text articles. We requested full texts via our institution’s interlibrary loan service if they were unavailable on the web. The first author screened the full texts for inclusion in consultation with the wider research team, and the research team verified the final list of included articles.

We used a pilot-tested form to record study characteristics and qualitative data on user experience (Multimedia Appendix 3). In total, 2 researchers independently charted data from a full text using a template adapted from the example evidence table for qualitative studies developed by the National Institute for Health and Care Excellence [33] and then met to discuss inconsistencies and improvements (Multimedia Appendix 4 [33]). The first author charted the remaining data. Our inclusive approach included qualitative data from any study type, such as qualitative data from qualitative studies (ie, studies that used a qualitative method of data collection and analysis), narrative data from qualitative components of mixed methods studies, and free text from questionnaire surveys as various types of qualitative evidence can enrich a synthesis [23]. In this study, charted qualitative data included quotations from participants and themes, theories, and interpretations generated by the studies’ authors. They were presented as narratives or summarized in tables and located in the Abstract, Results, and Discussion sections. We charted all qualitative data related to user experience as verbatim quotations. Multimedia Appendix 5 [34,35] provides a 17-page excerpt from our extensive data charting table.

We used the Quality of Reporting Tool [36] to critically appraise the included sources of evidence. The reporting of each study was appraised using 4 criteria: study design and research question, participant selection, data collection, and analysis. We assessed all qualitative studies overall (ie, as a whole) and all remaining papers (ie, mixed methods studies or questionnaire surveys) both overall and considering only qualitative data on user experience (ie, data included in our qualitative evidence synthesis). After pilot-testing the tool with 2 reviewers, a single reviewer categorized studies as “adequately reported” (satisfied at least 2 criteria) or “inadequately reported” (satisfied 1 or no criteria), and the first author verified all judgments and supporting evidence. These criteria have been used in other validated tools (eg, they represent items 3, 4, 5, and 8 from the Critical Appraisal Skills Programme qualitative checklist [37]) and in a review of barriers to and facilitators of engagement with digital mental health interventions [13].

As recommended in the Cochrane Handbook for Systematic Reviews of Interventions [23], we thematically synthesized charted qualitative data [38]. Thematic synthesis offers a clear and accessible inductive approach to produce descriptive themes that can evolve beyond the content of the primary studies into more in-depth analytic themes. The first author imported all charted qualitative data verbatim into the NVivo qualitative data analysis software (QSR International) and freely coded the data line by line according to their meaning and content using words directly from the data where possible. As qualitative evidence syntheses have received criticism for decontextualizing the findings of individual studies [38], the first author read all the charted data (including study aims, methods, and samples) before coding each study’s findings to preserve its original context and ensure that its findings could be fully understood without misinterpretation [39]. The first author then grouped similar codes into “descriptive themes” to summarize their meaning while keeping close to the original findings of the included studies. This was an iterative process that distilled users’ perspectives and experiences of using digital MBIs down to their key parts. In the next stage, the wider research team met to discuss the descriptive themes and develop “analytical themes,” which go beyond the findings of the primary studies by interpreting the key messages underlying the descriptive themes and using them to answer the review questions. We generated more abstract and analytical themes through an iterative process of inferring barriers, facilitators, and implications for intervention development from the descriptive themes and making changes to them where necessary. Multimedia Appendix 6 [23,38,39] provides more details about the analysis, including a 4-page excerpt from our list of codes, a full list of descriptive themes, and a comprehensive example of how we generated the analytical themes.

We took several steps to accelerate the review process so that evidence could be quickly incorporated into the initial phase of intervention planning [40]. First, we limited the inclusion criteria to English-language publications [25]. Second, we restricted the search to PsycINFO as an efficient way to achieve a manageable amount of relevant data (ie, by using a specialist database for psychological interventions [41] to retrieve studies most suitable for answering our review questions). This was necessary given that (1) too much data because of a large number of included studies can undermine qualitative evidence syntheses and (2) other methods of limiting the number of included studies are time and resource intensive (eg, purposive sampling [42]). Qualitative evidence syntheses aim to understand the phenomenon of interest in a context rather than aggregate data from large representative samples of studies to achieve statistical generalizability [42]; therefore, we do not anticipate this affecting the findings of this review. Third, one reviewer performed the full screening and data charting. We minimized the potential for increased errors and lower reproducibility because of this by piloting forms, estimating interrater reliability, and consulting with the wider research team. Multimedia Appendix 7 [25,38,40-46] provides more details on our streamlined approach.

The searches identified 530 unique records. Of these 530 records, 82 (15.5%) were included in the full-text review and 22 (4.2%) were included in the qualitative synthesis (Figure 1). We performed the first search on September 13, 2021, and reran the search on November 30, 2021, before analysis (Multimedia Appendix 8 [35,47-51]).

All studies were assessed as adequately reported (Multimedia Appendix 10 [16,34,35,47,48,52-68]), including qualitative studies (8/22, 36%) and mixed methods studies or questionnaire surveys when evaluated both as a whole and with respect to qualitative data on user experience only (14/22, 64%). Overall, each study reported on the study design and questions, participant selection, data collection, and analysis. When we evaluated mixed methods studies or questionnaire surveys considering only data included in our qualitative evidence synthesis, 32% (7/22) of the studies did not provide details of the analysis method (eg, the authors reviewed open responses for common themes without reference to or full description of the method), and 5% (1/22) of the studies did not describe data collection sufficiently.

We identified three themes: (1) responses to own practice, (2) making mindfulness a habit, and (3) leaning on others. Each theme is outlined in the following sections using illustrative quotes.

This review identified, critically appraised, and synthesized qualitative data from 22 original studies of people’s experiences using a digital MBI to identify factors that facilitate or act as barriers to their engagement with the intervention. Three overarching themes appeared to influence engagement: (1) responses to own practice, (2) making mindfulness a habit, and (3) leaning on others. Together, these themes provide crucial insights as to why people frequently stop engaging with digital MBIs. The following discussion elaborates on these areas and offers some recommendations for researchers and developers to guide intervention design and evaluation, thereby improving acceptability and engagement with digital MBIs, increasing their effectiveness, and supporting their translation to real-world use.

The first theme emphasized how adverse reactions to one’s own practice are common and may serve to reduce motivation. This suggests that the tendency to respond negatively to one’s own experience and application of mindfulness is a major barrier to using digital MBIs, which is consistent with the wider literature on mindfulness interventions and offers initial support for extending this finding to digital intervention formats. For example, in one study, the question “Am I doing it right?” emerged by the second week of a traditional MBCT course [69]. In another study, participants reported feeling self-critical when they could not make time to practice and when mindfulness did not appear to work for them [18]. As in this review, this negative reaction made it difficult for participants to continue to engage, prompting them to give up and remove it from their to-do list. To help overcome this barrier, traditional face-to-face programs such as MBCT explicitly allocate time to anticipating what difficulties and obstacles may arise in doing home practice (eg, trying to find free time) and how to deal with them [7]. Such content on overcoming barriers may be lost in the translation to digital formats, and our review is the first to highlight the importance of explicitly addressing this in digital MBIs.

This finding also indicates that one of the most important factors influencing engagement with digital MBIs is unique to mindfulness specifically rather than general to digital interventions and reflects the metacognitive nature of the intervention’s target behavior. Our review offers clear guidance on which particular combinations of factors identified across other literature (eg, on digital interventions or mindfulness interventions more broadly) are most influential in the specific context of digital MBIs, which is essential to make these interventions more persuasive, feasible, and relevant to users [20].

The second theme (making mindfulness a habit) highlighted the need and effort required to practice consistently and a call for personalization to help achieve this. This suggests that forming a mindfulness habit is a key barrier to sustained engagement with digital MBIs and that persuasive technological features could help overcome this barrier. Although prior work on digital interventions has identified personalization as an important feature, this review is the first to demonstrate its relevance to digital mindfulness interventions specifically. For example, a systematic review of web-based interventions found that the inclusion of persuasive design principles, including tailoring (ie, provision of content or feedback adapted to factors relevant to a user), explained 55% of the variance in session completion across studies [70]. Our findings suggest that certain factors that contribute to engagement with digital content in mobile and web-based interventions more generally may also apply to interventions for which engagement with the digital content is designed to facilitate completion of a nondigital target behavior (eg, an experiential mindfulness exercise) [11]. Notably, the threshold of engagement with the digital component that successfully facilitates the “non-digital target behaviour” can demonstrably vary between individuals [71], supporting a shift toward patient-treatment matching and person-centered care [72] and underscoring the need to implement this digitally (eg, through automated personalization).

Conversely, this theme diverges from the results of a thematic analysis of the experiences of health care professionals who participated in either a web-based or printed self-help MBI [18]. The health care professionals consistently reported that longer practices were more challenging to engage with than shorter practices, whereas our review found considerable variation in preferences for different intervention features (eg, format, materials, and sound), including length of practice, perhaps because of the breadth of MBIs included in our robust evidence synthesis. This highlights the importance of understanding the key behavioral and psychological needs of the target population to ensure that the intervention addresses them.

The third theme (leaning on others) highlighted that those engaging with digital MBIs are encouraged by additional support in its broadest sense (ie, any communication designed to support any aspect of the intervention, its completion, or its desired outcomes). This includes synchronous (eg, phone calls and web-based chats) and asynchronous (eg, email and SMS text messages) communication, support provided to a group of people (eg, discussion forums and group chats), and anything else (eg, automated reminders, technical assistance, feedback, and reaching out to someone). Although these results align with those of previous research on the impact of additional support in digital interventions [73], this study cannot draw conclusions on the relative power of each type of support because of the variability across studies. Given this, the provision of support in research settings needs to be considered. Interventions from almost all the studies in this review included additional support; however, it was not always clear what this constituted. For example, some studies (2/22, 9%) reported that participants could ask questions via email but did not specify whether they received clinical or purely technical assistance. Relatedly, participants may not have used the support on offer, although the results from this review indicate that this is not as important as having it available. Additional support in other studies (3/22, 14%) was provided to a group of participants; however, this type of support has been excluded from definitions of guidance [74]. Future research could explore whether there are unique barriers to engagement in guided versus unguided digital MBIs and compare different types and levels of support to advance understanding of how, when, and for whom additional support can improve engagement. This is important as there is a trade-off between the provision of support and scalability—if digital MBIs need to have someone always available to be engaging, they will be limited in reach and cost-effectiveness.

Irrespective of these uncertainties regarding the relative contributions of different types of support, it is worth noting that social support was found to be a key facilitator of engagement. This idea is consistent with the historical origins of mindfulness (ie, to be practiced collectively and in community [75]) and findings from in-person group settings. In a synthesis of the accounts of individuals with mental health difficulties in group MBIs [76], learning mindfulness within a group was found to be helpful as peer support encouraged perseverance with course demands and learning alongside people with similar experiences fostered a comfortable and destigmatizing environment. Our findings point to the idea that digital MBIs may suffer decreased engagement as a result of reduced social support.

Researchers can use the factors identified in this review to guide intervention design and, ultimately, improve engagement with digital MBIs. However, such strategies must be (1) grounded in relevant literature and (2) directly relevant to the individuals who will use the interventions. For example, the second theme suggests that instructing people to practice regularly is unlikely to turn it into a habit. Researchers might consider drawing on research on behavior change and habit formation, particularly with regard to digital interventions (eg, gamification technology to motivate behavior change). Researchers might also consider carrying out primary qualitative research to ensure that the generated strategies are informed by and meet the priorities and needs of the intended user. The person-based approach offers a systematic means of integrating theory, evidence, and user perspectives into initial intervention planning [19,20]. Therefore, the themes highlighted in this review could inform the production of guiding principles within this approach (ie, intervention design objectives and key features intended to achieve each aim).

To the best of our knowledge, this is the first review to synthesize qualitative evidence from individual studies across different contexts to advance the understanding of the barriers to and facilitators of engagement with digital MBIs. Using inductive thematic synthesis encouraged the generation of themes that “go beyond” the content of the primary studies to produce novel findings. All 22 studies were assessed as being adequately reported, which suggests that the papers included in this review are of sufficient quality to draw concrete inferences. We also followed established methodological guidance; used an a priori published protocol; and took several steps to increase the validity and reliability of the review, including pilot-testing forms and procedures, consultation with an information specialist, and regular team meetings.

In terms of limitations, we restricted our search to PsycINFO to manage the number of studies in a resource-efficient manner. However, it is possible that this led to the omission of additional relevant studies or introduced selection bias. Where possible (eg, in reviews with longer time frames), researchers should consider searching several sources and using purposive sampling to ensure that the final set of included studies meets relevant criteria (eg, has a wide geographic spread or rich data [42]). The studies included in this review reported mostly on White adult female individuals from Western countries, which means that the generalizability of our findings to underrepresented groups is unclear. This is an important area for further research as initial engagement with digital and mobile health interventions is lower in some underserved populations (eg, people of lower socioeconomic status [28]). Relatedly, we excluded studies with samples entirely aged <16 years and/or >35 years because of the focus of our own intervention development being on young people. Although the final age range covered was 16 to 69 years, future research would benefit from investigating engagement in younger and older populations as motivations to use digital interventions may vary.

There was significant heterogeneity across the interventions (eg, commercially available programs, acceptance and commitment therapy, mindful messaging, and guided mindfulness meditations) in the included studies, and these differences may have influenced engagement. Researchers and developers of digital MBIs should also consider how specific elements (eg, content, mode of delivery, and provision of support) might make people more or less likely to stop using the technology. Finally, although this review synthesized evidence from diverse study types, it is worth bearing in mind that engagement with MBIs is usually defined in terms of intervention use (ie, physical engagement [77]). It is unclear whether the factors identified in this review characterize facilitation and hindrance of aspects of psychological engagement, such as intention to practice mindfulness, belief that practicing mindfulness will be helpful, and commitment to integrating mindfulness into daily life. This is an important area for further research given evidence that psychological rather than physical disengagement from self-help MBIs has a greater impact on cultivating mindfulness [77].

Previous studies have shown the potential of digital MBIs to improve a range of health outcomes. Sufficient engagement with these interventions is required to achieve the intended effects; however, engagement is typically poor. This review synthesized evidence from studies on digital MBIs and identified 3 key factors that influence user engagement. We recommend that researchers generate their own solutions to these challenges by drawing on relevant literature and working with people from the target user population.