There were an estimated 4.1 million people living with and beyond breast cancer in 2022 [1]. This number is estimated to reach nearly 5 million by 2030 [2]. This population can encounter long-term challenges related to health and quality of life, including fatigue, anxiety and depressive symptoms, breast cancer–related lymphedema, metabolic dysregulation, bone loss and osteoporosis, and cancer recurrence [3,4]. Physical activity may protect people living with and beyond breast cancer from these problems [5-7]. However, most people who have been diagnosed with breast cancer do not meet nationally recommended physical activity guidelines [8,9].

Psychotherapy-informed digitally delivered acceptance- and mindfulness-based approaches may facilitate physical activity promotion for people living with and beyond breast cancer. People living with and beyond breast cancer commonly cite uncertainty, frustration, cancer-related fatigue, and pain to be barriers to physical activity [10-13]. Interventions that can help people to better navigate the uncomfortable thoughts and sensations that can act as impediments to physical activity may support their efforts to be more physically active. Authoritative entities recommend acceptance- and mindfulness-based interventions for people living with and beyond cancer; strong evidence supports their efficacy for reducing anxiety and depressive symptoms in this population, and they have been shown to reduce cancer-related fatigue [14-17]. Furthermore, physical activity interventions that simultaneously target both health-related behaviors and quality-of-life issues are more effective at achieving sustained physical activity outcomes than interventions that only promote physical activity for this population [18]. Emerging evidence suggests that acceptance- and mindfulness-based physical activity interventions can be feasible and effective at promoting physical activity [19-21] and that digitally delivered approaches can be acceptable and potentially effective for people living with and beyond cancer [22,23].

The degree to which participants meaningfully engage with and internalize health promotion content may determine the efficacy of digital behavior change interventions (DBCIs). However, achieving high levels of engagement can be a marked challenge for DBCIs [24]. Participants often exhibit low adherence to digital physical activity interventions and poor study retention [25]. Accordingly, researchers have tended to prioritize measuring and optimizing intervention use (eg, optimizing metrics such as the number of modules completed and time spent in an app) [26]. However, using system use data alone as a proxy for engagement is problematic. First, the relationship between these metrics and intervention efficacy is not always straightforward. Although there is sometimes a positive linear relationship between DBCI use and intervention efficacy, this is not always the case [25,27,28]. DBCIs may tend to have certain thresholds of engagement that confer the majority of the benefits, or intervention components may interact with one another. Furthermore, engagement with a DBCI may be expected to taper over time and even be supplanted by engaging in the desired behavior itself [29]. Second, focusing exclusively on quantitative use metrics affords only limited insight into the psychological aspects underpinning participants’ experiences with DBCIs. This precludes achieving a full understanding of the DBCI’s mechanisms of action. Thus, researchers have called for broadening our conceptualization of DBCI engagement [28].

A growing literature highlights the importance of understanding the psychological processes that occur when participants engage with DBCI content; this includes supplementing system use metrics with additional aspects of DBCI engagement that pertain to affect, attention, interest, immersion, flow, and reflection [26,29,30]. Researchers of human-computer interaction have discussed engagement in similar terms, parsing engagement according to notions of behavioral adherence, behavioral effort (eg, discussing emotions, thoughts, and behaviors in intervention), cognition, and affect [31]. Researchers with an educational perspective have provided similar definitions of engagement with digital systems, noting that the cognitive and affective aspects of engagement are largely neglected [32]. Yang et al [25] have contributed a framework specific to physical activity promotion in mobile health in which they conceptualize engagement as being determined by breadth, depth, interaction, and length of engagement. These broad conceptualizations are useful for helping to orient researchers toward the many and varied facets of engagement. Rather than attempting to establish a universal conceptualization, it may be advantageous to tailor the conceptualizations of engagement based on context [33]. Using qualitative and mixed methods to investigate the nature of participants’ engagement with DBCIs may be particularly useful to this end [24,26,29].

Reflection is an essential process for self-improvement and making lasting behavior changes in the context of DBCIs [34]. As applied to health-related education, reflection and reflective processes connotate deliberate critical analysis of knowledge and experiences to achieve a fuller understanding [35]. The process of reflecting on the outcomes of past and ongoing efforts or newly acquired information helps individuals integrate knowledge and skills into practice and overcome persistent barriers to behavior change [35,36]. Reflection may be particularly relevant in the context of DBCIs because it may allow people to gain meaningful insights from their personal health-related data that can support lasting behavior change [34]. In the education literature, reflective thinking and writing are commonly used strategies to prompt personal reflection, and it is a common practice to evaluate written responses to determine to what degree an individual has considered and applied didactic content in the context of their own lived experiences [37-39]. Fleck and Fitzpatrick [40] present a framework that operationally defines 5 levels of reflection. This framework has been shown to be particularly useful for evaluating the levels of reflection promoted by the design features of DBCIs [34]. In this study, we sought to extend this literature by characterizing the reflective processes engaged in by people living with and beyond cancer who experienced a digital acceptance- and mindfulness-based physical activity intervention.

It may be possible to help people living with and beyond breast cancer reframe some of the unpleasant internal sensations that can act as barriers to physical activity (eg, pain, fatigue, and frustration). Understanding the degree to which participants reflect on and internalize digital intervention content targeting these psychological processes may be useful for understanding how best to support this population. This study answers calls to investigate the deeper psychological aspects involved in engagement with DBCIs to provide insight into the interplay among intervention use, individual experiences, and intervention efficacy. The aims of this study were to (1) characterize the nature of reflective processes brought about by a DBCI designed to increase physical activity in insufficiently active people living with and beyond breast cancer and (2) explore the potential utility of a metric of reflective processes in DBCIs within an acceptance- and mindfulness-based physical activity intervention.

We conducted a secondary analysis on data obtained from a 1-group pilot study. The purpose of the parent study was to evaluate the acceptability of an acceptance- and mindfulness-based intervention to increase physical activity in survivors of breast cancer [23]. This study is an investigation of the use of the intervention. Participants were female adults (aged ≥18 y) who had been diagnosed with breast cancer but were not actively preparing for surgery or undergoing chemotherapy or irradiation treatment (n=75). Eligibility criteria included that participants reported engaging in <150 minutes of moderate-intensity aerobic physical activity per week. We recruited participants using a large listserve of individuals who were interested in receiving information about breast cancer–related research studies. Study staff contacted interested individuals via telephone to assess eligibility.

Study procedures were approved by the University of Texas School of Public Health Committee for the Protection of Human Subjects (HSC-SPH-18-1025). All participants provided informed consent for participation.

The intervention has been described in detail elsewhere [23]. Briefly, the ACTive program was an 8-week DBCI designed to help insufficiently active survivors of breast cancer increase moderate-intensity aerobic and muscle-strengthening physical activity. It was centered on increasing physical activity acceptance (ie, cognitive acceptance and behavioral commitment) and autonomous motivation for physical activity (ie, enjoyment, values, interest, and identification). It was grounded in acceptance and commitment therapy (ACT) principles (ie, values, committed action, acceptance, defusion, and contacting the present moment). It consisted of 8 modules (along with a brief introductory module), each of which was delivered electronically via a weekly REDCap (Research Electronic Data Capture; Vanderbilt University) survey. The modules included didactic content and experiential exercises targeting core ACT principles. They were presented in brief videos, audio files, images, and other documents. Participants were prompted to input information periodically (eg, physical activity levels and reactions and thoughts on experiential exercises). Branching logic within REDCap was used to remind participants of their responses to items from previous modules as well as provide feedback on their responses within modules (eg, an emoji appeared if responses indicated that the participant had met her physical activity goal for the week). The modules contained a repository of additional optional content (eg, muscle-strengthening physical activity videos for survivors of cancer, videos with yoga classes for survivors of breast cancer, and a video on proper walking posture for survivors of cancer).

At the end of modules 1 to 7, an item asked participants, “What is one important, personal take-away point from this session?” This item was included to identify specific aspects of the intervention that were perceived as the most useful by participants because usefulness is an important component of acceptability. Participants were presented with an open-text box to provide a reply. Providing participants with the opportunity to reply to open-ended items such as this can help contribute to a more in-depth understanding of participants’ engagement and experience with DBCI content [26].

Reflection levels were based on the reflection framework presented by Fleck and Fitzpatrick [40]. This framework emerged from research conducted in the context of the design of digital technologies and human-computer interaction. It defines 5 levels of reflection intended to serve as a resource for thinking about, and designing for, reflection. Reflection level 0 (revisiting) is defined as “Description or statement about events without further elaboration or explanation. Not reflective.” Reflection level 1 (reflective description) is defined as “Description including justification or reasons for action or interpretation, but in a reporting or descriptive way. No alternate explanations explored, limited analysis and no change of perspective.” Reflection level 2 (dialogic reflection: exploring relationships) is defined as “A different level of thinking about intervention content. Identifying or exploring relationships between relevant concepts. Applying experience or knowledge, providing evidence of cycles of interpreting and questioning, consideration of different explanations, hypotheses and/or other points of view.” Reflection level 3 (transformative reflection: fundamental change) is defined as “Revisiting an event or knowledge with intent to re-organize and/or do something differently. Asking fundamental questions and challenging personal assumptions leading to a change in practice or understanding.” Finally, reflection level 4 (critical reflection: wider implications) is defined as “Social and ethical issues are taken into consideration. Generally considering the (much wider) picture.”

The average age of the participants was 57.2 (SD 11.2; range 31-78) years. The median time since breast cancer diagnosis was 8.2 (IQR 3-12) years. The participants were relatively well educated, mostly non-Hispanic White (58/75, 77%), and mostly overweight or obese (54/75, 72%; Table 1).

The median number of the 8 modules completed was 8 and the first quartile value was 4 modules (IQR 4). The minimum number of modules completed was 1 (7/75, 9%).

The coders created a table while going through the process of coding the reflections based on the reflection framework presented by Fleck and Fitzpatrick [40]. We added descriptors and illustrative examples from this study to further clarify our conception of the various reflective levels as they might be applied in the context of DBCIs (Table 2).

We observed associations between intervention adherence and change in 2 outcomes of interest (Table 3). Higher intervention adherence was associated with higher weekly bouts of muscle-strengthening physical activity at follow-up, adjusting for baseline levels, age, time since diagnosis, stage at diagnosis, race, ethnicity, education level, BMI category, and mean reflection level (B=0.26, SE 0.12, 95% CI 0.02-0.50; Cohen d=2.17). Participants in the fourth quartile of intervention adherence (21/75, 28%) averaged an increase of 0.38 bouts of muscle-strengthening physical activity in a typical week from baseline to follow-up, whereas those in the first quartile (40/75, 53%) averaged an increase of 1.43 bouts in a typical week. Similarly, higher intervention adherence was associated with less sleep disturbance at follow-up, adjusting for baseline levels and other covariates (B=−1.04, SE 0.50, 95% CI −2.02 to −0.06; Cohen d=2.08). Participants in the fourth quartile of intervention adherence averaged a decrease of 2.31 in the PROMIS-29 sleep disturbance score from baseline to follow-up, whereas those in the first quartile averaged a decrease of 4.68 in the PROMIS-29 sleep disturbance score. We did not observe statistically significant associations between intervention adherence and other outcomes of interest.

We observed associations between participants’ mean reflection level and change in 2 outcomes of interest (Table 3). Higher mean reflection level was associated with higher integrated regulation at follow-up, adjusting for baseline levels, age, time since diagnosis, stage at diagnosis, race, ethnicity, education level, BMI category, and intervention adherence (B=0.55, SE 0.25, 95% CI 0.06-1.04; Cohen d=2.20). Similarly, higher mean reflection level was associated with higher cognitive acceptance of physical activity at follow-up, adjusting for baseline levels and all other covariates (B=3.42, SE 1.70, 95% CI 0.09-6.75; Cohen d=2.01). We did not observe statistically significant associations between mean reflection level and other outcomes of interest.

This study was a secondary analysis of data from a digitally mediated intervention centered on applying principles and techniques from ACT to help people living with and beyond breast cancer overcome internal barriers to physical activity (eg, pain, fatigue, and frustration). We applied the framework presented by Fleck and Fitzpatrick [40] to gauge the depth of reflection evident in participants’ written responses to prompts encouraging reflection on the intervention’s weekly modules. We conducted content analysis to characterize participants’ responses and multiple linear regression analyses to explore to what extent intervention adherence and participants’ mean level of reflection were associated with study outcomes (ie, physical activity behaviors, motivation for physical activity, physical activity acceptance, and health-related quality of life). There were 407 written responses from the 75 participants over the course of 7 modules. Most of the responses were rated as either reflection level 0 (ie, description; 70/407, 17.2%) or level 1 (ie, reflective description; 247/407, 60.7%). Of the 407 responses, 90 (22.1%) demonstrated evidence of deeper levels of reflection. Intervention adherence was associated with more muscle-strengthening physical activity and better sleep outcomes. Mean reflection level was associated with more integrated motivation for physical activity and higher willingness to experience the full range of sensations that may accompany physical activity.

This study extends previous literature that has highlighted the importance of the cognitive and experiential aspects of DBCI engagement. To date, these aspects of DBCI engagement have largely been inferred from behavioral data (ie, system use metrics). In their review of methodologies for measuring engagement with DBCIs, Short et al [26] suggested that researchers should investigate how intervention content affects the cognitive and experiential aspects of DBCI engagement via the inclusion of open-ended items. We took this approach and found that participants’ responses to open-ended items could be evaluated using the reflection framework presented by Fleck and Fitzpatrick [40]. This approach is consistent with research in the education literature that has found reflective writing samples to be amenable to the quantitative assessment of the depth of reflection [35].

Our findings regarding the breakdown of reflection levels evident in this study were consistent with reviews that have evaluated reflective writing in the context of health-related education [39,47]. Even among graduate students, the great majority of reflective processes tend to occur at the descriptive level, and transformative and critical reflections are consistently rare [39,47]. There is limited literature applying these or similar techniques to better understand programs for health education and health promotion [39,47]. In this study, we observed considerable variation in the reflective depth of responses within and between individuals. This suggests that depth of reflection may be modifiable in this context. Indeed, reflection is a common goal of educational interventions, and it seems to be modifiable in other contexts [35]. Providing appropriate scaffolding, fostering collaboration in learning, and using varied exercises to stimulate reflection have been identified as techniques that may increase learners’ depth of reflection [35,39]. Time constraints, conflicting values, a lack of feedback, and a lack of trust have been identified as barriers to learning in the context of educational interventions promoting student reflection [35,39]. Future research is needed to identify how DBCIs might be designed to facilitate meaningful reflection and most effectively target the psychosocial determinants of lasting behavior change.

The themes identified in our qualitative analysis provided some insight into the variability of cognitive processes occurring in participants who experienced the digital physical activity intervention. Common themes in the lower levels of reflection generally suggested that participants had acquired new knowledge and endorsed intentions to change their physical activity behaviors (eg, making a resolution and knowledge of subject matter). These are important antecedents of successful behavior change [48]. However, they are generally not sufficient for realizing physical activity adherence goals; nearly half of the people who indicate that they intend to change their physical activity patterns do not do so [49]. The themes that characterized higher levels of reflective comments are concordant with processes that have been shown to moderate this intention-behavior gap [50]. Discussion of personal facilitators or mechanisms of behavior change and experiencing personal insight or a fundamental change in perspective conceptually align with the self-regulatory processes and notions of physical activity identity that tend to moderate the relationship between intention and physical activity adherence [50]. Encouraging personal reflection may be an autonomy-supportive approach to health promotion [34,51,52]. More research is needed to better understand if and how engaging in deep reflection may influence the conscious (eg, affective attitude) and automatic (eg, identity) processes that underlie successful behavior change.

Reflection may be a type of cognitive engagement uniquely suited to supporting knowledge transfer into other domains and contexts [36,53,54]. Schon [55] provides detailed commentary on reflective processes and distinguishes reflection-on-action (critical retrospective analysis) from reflection-in-action (conscious awareness of real-time behavioral modification). The author emphasizes the primacy of the latter as a determinant of sustained change and suggests that reflection-on-action may serve as a prerequisite for implementing change in real time. We found participants’ average depth of retrospective reflection to be associated with more integrated regulation for physical activity and cognitive acceptance in the context of physical activity but not with other outcomes of interest. It is encouraging that deeper levels of reflection on intervention content were associated with beneficial changes in motivation and physical activity acceptance, given that these were the theory-informed psychosocial constructs that the DBCI targeted [23]. The cognitive acceptance of physical activity has been shown to be associated with long-term changes in objectively measured physical activity, and integrated regulation is similarly predictive of physical activity [45,56]. However, in this study, we did not observe evidence that the depth of reflection was associated with physical activity–related outcomes or other outcomes of interest. Simple intervention adherence was associated with increases in muscle-strengthening physical activity and reductions in sleep disturbance. This may be due to the intervention’s inclusion of practical resources for muscle-strengthening physical activity and mindfulness exercises. Our findings suggest that designing DBCIs to encourage reflective processes may help target some theory-supported mechanisms of action but may not always be necessary to engender changes in desired end points. More research is needed to investigate whether deep reflection mediates changes in key long-term behavioral and health outcomes and to what degree it should be prioritized in DBCIs.

Considering the depth of reflection may be a useful lens through which to evaluate participants’ cognitive engagement with the didactic components of DBCIs. At present, DBCIs are commonly oriented toward maximizing participant engagement as measured by system use. However, rather than simply attempting to maximize DBCI system use, it may be beneficial to target aspects of effective engagement [29]. In this study, participants’ depth of reflection was associated with 2 key study outcomes. It may be possible to design dynamic interventions that optimize psychological processes such as critical reflection. This endeavor would be supported by the ability to derive metrics that reliably reflect underlying psychological processes from participants’ verbal output. We demonstrated that applying the reflection framework presented by Fleck and Fitzpatrick [40] is a promising approach to quantifying participants’ qualitative data. Emerging technologies such as large language models may be applied to this end to make this process more expedient and conducive to just-in-time adaptive interventions. Digital health promotion efforts centered on optimizing psychological processes such as critical reflection may supplement or supplant approaches that are narrowly oriented toward maximizing system use.

Reflective writing can serve as a means of self-expression and have therapeutic effects [57]. However, it can also be perceived as burdensome by some individuals. Perski et al [30] emphasized the role played by participants’ subjective experiences, characterized by attention, interest, and affect, in DBCI engagement. There are likely important trade-offs that occur in optimizing for cognitive versus affective aspects of DBCI engagement. Participants’ comments in this study suggested that although many enjoyed the introspective aspects of the intervention, some did not. This observation is concordant with high ratings of acceptability in the parent study for those who completed the study but a less-than-ideal dropout rate (23.7%) [23]. Achieving deeper levels of reflection may have a dynamic and, in some contexts, diametrically opposed relationship with affective or subjective experiences. Particularly given the importance of affect in physical activity behavior [58], future research should investigate the interrelationships among these aspects of DBCI engagement and how to strike the right balance for different individuals in different contexts.

The findings from this study must be considered in the context of its limitations. A small sample size, possible selection bias, and high attrition limit the generalizability of the findings of this study; the analytical sample was a convenience sample that was relatively well educated and had limited racial and ethnic diversity. Short et al [26] highlight that the use of qualitative methods to assess DBCI engagement is inherently limited by a lack of generalizability. The application of the reflection framework presented by Fleck and Fitzpatrick [40] along with directed content analytic methods may facilitate comparison in future studies; however, it is important to note that there may be other dimensions of reflection. We were primarily interested in assessing the vertical dimension of reflection (ie, the depth of reflection), but reflection has also been characterized as having an iterative process-oriented dimension that can be nonlinear and cyclic [39]. Given this prospect and the sometimes cyclical nature of behavior change itself [59], it may be beneficial to longitudinally investigate reflective processes and their bidirectional relationships with physical activity behaviors and determinants. Likewise, there are other ways to define intervention adherence. A limitation of this study is that we did not obtain other metrics that might characterize adherence, such as time spent in the modules. In future studies, obtaining more granular data concerning system use may be advantageous. Given that the study was conducted during the COVID-19 pandemic, history was a threat that may have influenced participants’ reflective processes and physical activity–related constructs. Reliance on the self-reported assessment of physical activity patterns also has well-documented limitations. Finally, we conducted multiple statistical tests and, although this investigation was explicitly exploratory in nature, the study findings are prone to an inflated chance of type I error.

In this study, we sought to explore a novel method for understanding critical reflection occurring in a DBCI designed to increase physical activity in insufficiently active people living with and beyond breast cancer. We found the application of qualitative content analysis based on the reflection framework presented by Fleck and Fitzpatrick [40] to be a useful tool for helping to gauge the extent to which participants engaged in reflective processes. Furthermore, we found that deeper reflection levels tended to be associated with changes in the targeted psychosocial constructs. Reflecting on newly acquired information is a critical process in integrating relevant insights for sustained behavior change. Encouraging personal reflection is an autonomy-supportive approach to promoting physical activity, and more research is warranted to investigate this approach in DBCIs serving people living with and beyond cancer.