My Best Self in the Past or Future: A Randomized Controlled Trial Examining Adherence, Engagement, Age and Mental Health in a Mobile-based BPS Intervention
- Open Access
- 01.10.2025
- Original Research
Erschienen in:
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by (Link öffnet in neuem Fenster)
Abstract
1 Introduction
A widely researched beneficial psychology activity is imagining one’s ideal future self. The Best Possible Self (BPS) is a positive future thinking technique developed by King (2001). Typically, participants are asked to project themselves into the future and image that everything has worked out in the best possible way. Possible selves are one’s ideas about how they will become in the future, and this can clarify one’s goals, aspirations and motives and act as an incentive for future behaviors (Markus & Nurius, 1986; Sheldon & Lyubomirsky, 2006).
Different meta-analyses have shown the mental health benefits of the BPS intervention, as well its robustness across a variety of delivery methods. A meta-analysis of the effectiveness of BPS interventions on optimism showed that the intervention was more effective than other optimism interventions in increasing optimism levels (Malouff & Schutte, 2017). Other meta-analyses found that the BPS intervention had non-significant effects on negative affect and depressive symptoms, but significantly increased positive affect and optimism (Carrillo et al., 2019; Heekerens & Eid, 2021; Schubert et al., 2020). The robustness of the BPS intervention has been further supported by a meta-analysis indicating that a variety in its components – whether solely writing based (e.g. King, 2001) or combining writing and imagery components (e.g. Boselie et al.,. 2023) – did not influence effectiveness of the BPS intervention (Carrillo et al., 2019). Likewise, the delivery method, whether conducted individually (e.g., Meevissen et al., 2011), in small groups (e.g., Peters et al., 2010), face-to-face (e.g., Layous et al., 2013) or online (e.g., Lyubomirsky et al., 2011), did not influence effectiveness of the BPS intervention either (Carrillo et al., 2019). In line with these outcomes, a qualitative review of BPS interventions concluded that BPS interventions are a viable and adaptive strategy for improving mental health across a variety of populations and delivery methods (Loveday et al., 2018). Nowadays, the online delivery of the BPS intervention is the standard-delivery method as it offers several advantages. It gives the possibility to deliver the BPS anonymously, at low cost, and to very large populations (Mitchell et al., 2010).
Anzeige
Independent of delivery method, several mechanisms have been proposed to account for the effectiveness of the BPS. One of these mechanisms may be shaping one’s narrative identity. The BPS exercise may facilitate the internalization and development of agentic and positive self-narratives, thereby fostering a greater sense of coherence (McAdams, 2008). Additionally, accessing positive self-relevant information may further contribute to BPS effectiveness, as thinking positively about oneself has proven to have beneficial effects (Mongrain et al., 2012). The temporal frame of the BPS may also be an important factor to explain its effectiveness, as time – past, present and future – is fundamental to various aspects of wellbeing (Durayappah, 2011). A positive outlook on the past requires emotion regulation after an event through positive reminiscence. This process entails recalling favorite memories, vividly replaying these memories and the cultivation of gratitude (Quoidbach et al., 2015), which has been associated with increased life satisfaction (Przepiorka, & Sobol-Kwapinska, 2021). On the other hand, a positive future-oriented outlook involves anticipatory emotion regulation. This requires the visualization of desired outcomes, envisioning the behaviors necessary to achieve these outcomes and anticipating the associated emotional rewards (Quoidbach et al., 2015). This positive future-oriented perspective promotes adaptive behaviors such as healthy coping strategies and increases hope, self-efficacy and a positive life-style (Kooij et al., 2018). A previous study (Carrillo et al., 2021) examined the effect of the temporal frame in online BPS interventions. Participants were instructed to focus on their Best Past Self, their Best Present Self or on their BPS. All these time frames increased well-being, but there were no differences in effectiveness between the three conditions, nor were any of the conditions associated with specific temporal variables, such as past life satisfaction. The authors concluded that other processes other than temporal focus might be relevant in the effectiveness of the BPS (Carrillo et al., 2021).
We propose that adherence and engagement – both key predictors of online intervention outcomes (Beatty & Binnion, 2016) - are critical factors in the effectiveness of the BPS intervention. To date, these factors have not been examined in relation to BPS efficacy. However, their influence may be particularly salient when considering temporal frame and individual characteristics, such as age.
1.1 Adherence and Engagement
Adherence has been defined as the extent to which individuals experience the content of the online intervention as intended by the creator (Christensen et al., 2009; Kelders et al., 2012). It differs from dropout, which refers to individuals who fail to complete the research protocol, for instance, by not completing questionnaires (Christensen et al., 2009). Adherence can be assessed by comparing an individual’s observed usage to the intended usage of the intervention (Kelders et al., 2012). Although adherence to online interventions has a beneficial effect on mental health outcomes (e.g., Roepke et al., 2015), the assumption that increased frequency of use equates to “better results” has been challenged, as adherence does not consistently predict intervention efficacy (Schlosser et al., 2018; Graham et al., 2020).
Engagement is commonly understood as the degree to which an individual is involved or absorbed in something (Kelders, 2020a). It is a subjective experience characterized by attention, interest, and affect (Perski et al., 2017). Engagement is different from adherence, for example, believing that an app is very helpful in achieving one’s goal, does not imply that a person will adhere to the intended use of, for instance, once a day (Kelders et al., 2020b). The Model of User Engagement (Short et al., 2015) proposes that environmental factors, individual characteristics, and intervention features interact to influence user engagement. Environmental factors, such as time availability and internet access, indirectly shape engagement by affecting user expectations and usability. Individual characteristics, particularly the perceived personal relevance of the intervention, also impact motivation to engage. Intervention features—including aesthetics, frequency of use, delivery mode, and content—directly influence engagement. These elements interact dynamically, influencing each other and engagement itself (Short et al., 2015).
Anzeige
A systematic review showed that engagement was related to intervention effectiveness (Perski et al., 2017). For instance, individuals who perceived a digital mental health intervention as more useful and easier to use and learn, had greater improvements in depression and anxiety over eight weeks (Graham et al., 2021). Notably, in this study, adherence, reflected in the number of app sessions, was not significantly related to outcomes in this study (Graham et al., 2020). This suggest that engagement may be as important as, or even more critical than adherence in eliciting mental health improvements.
Age may be an important individual characteristic that influence engagement (Short et al., 2015). It has been related to engagement, but the association between age and engagement was not clear, as age was both positively and negatively related to engagement in a meta-analysis (Borghouts et al., 2021). Perhaps the association between age and engagement depends on specific factors of the online intervention, such as the content and the objective of the application. The BPAS might be more personal relevant for older participants because attention to the past increases with age (Laureiro-Martinez et al., 2017), whereas the BPS can be more effective for participants at a younger age because younger people still experience more opportunities to reach their goals compared to older people (Lyubomirsky & Layous, 2013; Meevissen et al., 2011). Hence, age might influence the association between the temporal frame of the BPS intervention, engagement and BPS efficacy.
1.2 The Current Study
The aim of this study is to examine the specific and generic mechanisms of change in an BPS intervention delivered via a mobile application. Temporal frame, adherence, engagement and age will be related to BPS intervention efficacy. Four hypotheses were formulated: (1) visualizing one’s best self in the past (BPAS) or future (BPS) will enhance well-being to a higher extent than the control condition (Carrillo et al., 2021); (2) engagement will be a stronger predictor of BPS efficacy than adherence, irrespective of condition (Graham et al., 2020, 2021; Perski et al., 2017); (3) visualizing one’s best self in the past (BPAS) or future (BPS) will be more engaging than the control condition and this will relate to BPS efficacy (Short et al., 2015); (4) engagement mediates the association between condition and BPS efficacy and these associations will be influenced by age. The association between BPAS, engagement and BPS efficacy will be stronger for participants with a relative older age (Borghouts et al., 2021; Laureiro-Martinez et al., 2017; Lyubomirsky & Layous, 2013; Meevissen et al., 2011).
2 Method
2.1 Study Design
The study was designed as a randomized controlled trial (RCT) in three parallel groups. Participants were randomized into one of three conditions: (1) Best Possible Self (BPS), (2) Best Past Self and Best Possible Self (BPAS/BPS), or a (3) control condition. Online measurements were taken at baseline (T0), one week after the start of the intervention at mid-intervention (T1), two weeks after baseline immediately after the intervention at post-intervention (T2) and six weeks after baseline at the four-week follow-up (T3). This follow-up period was determined based on prior research in which participants practiced the BPS intervention daily over a two-week period, with significant effects assessed at a two-week follow-up (Meevissen et al., 2011). Extending the follow-up to four weeks allowed for an examination of whether these effects persisted beyond two weeks. Furthermore, research suggest that while the benefits of the BPS intervention can be maintained up to one-month post-intervention, they tend to diminish over time, with effects no longer reaching significance at three months (Enrique et al., 2018) or six months (Lyubomirsky et al., 2011).
The study was approved by the Faculty of Behavioural Sciences Ethics Committee at the University of Twente, under registration number BCE16337, and registered in the United States National Institute of Health Registration System (NCT03024853).
2.2 Participants and Procedure
The sample size needed was determined using G*Power 3.1 (Faul et al., 2009). A total of 215 participants needed to be included in the sample to detect an effect size of Cohen’s d = 0.27, which is the mean of effect sizes found in the most recent meta-analyses of the BPS intervention on the primary outcome (Carrillo et al., 2019; Heekerens & Eid, 2021), an alpha error of 0.05 and a statistical power of 1–β = 0.95. Based on the dropout rate of the study by Carrillo et al. (2021), which was around 60%, a minimum number of 344 participants needed to be included.
Participants were recruited from the general population via advertisements on Facebook. The recruitment message was: ‘Do you want to grow your confidence? Soon we will start with a study in which you will boost your confidence, happiness and satisfaction with life through exercises presented on a mobile application for a two-week period’. In this way a self-selected ‘well-being seeking’ sample was recruited. The advertisement contained a link to the research webpage where the study purpose was explained in more detail and where people could sign up by completing an online screening questionnaire.
Participants were eligible for participation if they: (a) were 18 years or older; (b) had a smartphone with internet connection; (c) owned an email address; (d) had sufficient proficiency of the Dutch language; (e) did not report severe levels of anxiety as indicated by a score ≤ 15 on the GAD-7 (Spitzer et al., 2006); (f) did not report severe depressive symptoms, as determined by a score ≤ 34 on the CES-D (Thomas et al., 2001), (g) had low to moderate levels of well-being, as determined by the MHC-SF (see section Measures), and (h) provided informed consent. Participants that were excluded based on the scores on the GAD-7 and CES-D were advised to contact their general practitioner if they felt distressed by their symptoms or when symptoms persisted for a prolonged period.
After passing the initial screening, participants received an email invitation with a link to the first online questionnaire, which, along with the screening, served as the baseline measurement. Upon completing the questionnaire, participants were instructed to download the mobile app. Those who did were randomly assigned to one of three conditions using an unrestricted randomization process through www.random.org, ensuring equal probability for each participant. A research assistant conducted participant allocation, and none of the researchers were aware of individual allocation to conditions.
To minimize drop out from the study, different strategies were used. Email reminders were sent for completing questionnaires. Participants who completed all measurements were entered into a raffle for one 100-euro gift card, five 50-euro gift cards, and twenty 10-euro gift cards.
2.3 Intervention
2.3.1 Mobile App
A specific mobile app was created in which participants could practice the intervention along two weeks and complete the mid-term, post- and follow-up measurements. In this app, an avatar called ‘Dan’ guided the intervention (see Fig. 1). He gave details about the intervention, its duration and how it worked. He explained that participants would practice their imagination every day and that the exercises would take about five to ten minutes. Next, Dan introduced a practice visualization exercise in which participants had to visualize cutting, smelling and biting a lemon, activating all their senses. After this practice exercise, he introduced the imagination exercise depending on the allocation to the different conditions. During the rest of the intervention, Dan reminded participants to practice. Every imagination exercise, regardless of the conditions, started with the instructions to find a quiet place where participants would not be disturbed, to sit straight, to close their eyes and follow their breathing. The exercises were audio-guided by Dan, so participants could close their eyes and follow the audio instructions.
Fig. 1
Guide Dan
To promote adherence, Dan gave psychoeducation about distractions during the two weeks intervention with short messages such as “a wandering mind is normal during the imagination exercise” or “practice would make the exercise easier”. Additionally, after completing the imagination exercise, Dan complimented the participants and showed participants the progress of the training.
2.3.2 Best Possible Self Condition
The BPS condition consists of visualizing oneself in the future after everything has gone as well as it possibly could (King, 2001; Meevissen et al., 2011; Sheldon & Lyubomirsky, 2006). In the course of two weeks participants were instructed to imagine their best possible self with regard to their personal strengths, social relationships, professional achievements, during leisure time and to imagine their best possible self in every domain. These small differences between the exercises were based on previous research (Meevissen et al., 2011) and were added along the intervention to provide variety and ensure that participants did not have to do exactly the same exercise for two weeks.
2.3.3 Best Past Self/Best Possible Self Condition
In the first week of the BPAS/BPS condition, participants were instructed to recall and visualize themselves in a time in the past when they felt they displayed the best version of oneself, focusing on the goals they had achieved and the best features they had at that time (Carrillo et al., 2021). After one week, the instructions changed and participants were asked to visualize their best possible self. In both weeks, the same variances in exercises were provided as in the BPS condition.
2.3.4 Control Condition
In the control condition, participants had to visualize the activities they did during the last 24 h (Carrillo et al., 2021; Meevissen et al., 2011; Sheldon & Lyubomirsky, 2006). To provide variations in exercises, participants were instructed to focus on the activities they did in the morning, afternoon, evening, or all activities in the last 24 h.
2.4 Measures
2.4.1 Screening
Depression was measured with the Center for Epidemiologic Studies Depression scale (CES-D; Radloff, 1977). It is a widely used 20-item instrument to assess depression over the past week. Responses are rated on a 4-point scale from 0 = “rarely or none of the time (less than one day)” to 4 = “most or all of the time (5–7 days)”. An example item is “I felt depressed”. The internal consistency was good (Cronbach’s α: = 0.81).
The Generalized Anxiety Disorder-7 (GAD-7; Spitzer et al., 2006) is a 7-item self-report anxiety questionnaire that assesses the severity of anxiety symptoms over the last two weeks. An example item is: “Over the last 2 weeks, how often have you been bothered by the following problems? Feeling nervous, anxious or on edge”. Items were rated on a 4-point scale from 0 = “not at all” to 3 = “Nearly every day”. In this study, the internal consistency was acceptable (Cronbach’s α: 0.76).
Well-being was measured using the 14-item Mental Health Continuum – Short Form (MHC-SF; Keyes, 2006) comprising three subscales: emotional (three items), psychological (six items) and social well-being (five items). Items (e.g. “In the past month, how often did you feel happy?”), were rated on a 6-point scale (1 = never; 6 = every day). Low to moderate mental health was defined as ≤ 4 on the emotional well-being scale and less than five scores ≥ 5 on the combined psychological and social well-being scale (Keyes, 2006). The internal consistency in this study was good (Cronbach’s α = 0.81).
2.4.2 Primary Outcome
Dispositional optimism was measured using the Life Orientation Test-Revised (LOT-R, Scheier et al., 1994), selected for its sensitivity to change in online BPS interventions (e.g. Boselie et al., 2023). The scale consists of 10 items (e.g. “I usually expect the best”), including four filler items, rated on a 5-point scale (1 = strongly disagree; 5 = strongly agree) In this study, internal consistency was acceptable across all time points (Cronbach’s α = 0.77 − 0.82).
2.4.3 Secondary Outcomes
Affect was measured with the 20-item Positive and Negative Affect Scale (PANAS; Watson et al., 1988), widely used in digital mental health research and sensitive to change (e.g. Carrillo et al., 2019). Participants indicated on a 5-pont scale (1 = not at all; 5 = extremely) how strongly they generally felt emotions such as “Enthusiastic” (Positive Affect, PA) or “Upset” (Negative Affect, NA). Each subscale included 10 items, with good internal consistency across measurement points (Cronbach’s α: PA = 0.84 − 0.89; NA = 0.84 − 0.87).
The Temporal Satisfaction with Life Scale (TSWLS; Pavot et al., 1998) assesses past, present, and future life satisfaction with 15 items (five per subscale). The TSWLS has shown adaptability and reliability in online mental health interventions (e.g., Carrillo et al., 2021). Items (e.g., “I am satisfied with my life in the past”) were rated on a 7-point scale (1 = strongly disagree; 7 = strongly agree). Internal consistency was good across all time points (Cronbach’s α: past life satisfaction = 0.85 − 0.90; present life satisfaction = 0.88 − 0.92; future life satisfaction = 0.82 − 0.89).
2.4.4 General Mechanisms of the Intervention Effect
Adherence was measured as the number of times participants used the TIIM app during the two-week intervention period. For each day, participants were scored “0” if they did not use the app or “1” if they did. Adherence at T1 and T2 ranged from 0 to 7, total scores ranged from 0 to 14, with higher scores indicating higher levels of adherence.
Engagement was measured using the 9-item Twente Engagement with Ehealth Technologies Scale (TWEETS; Kelders et al., 2020b). The scale measures behavioral, cognitive, and affective engagement with health applications, using items such as “I enjoy using this app”. Items were rated on a 5-point scale (1 = strongly disagree; 5 = strongly agree). Both at T1 and T2 the questionnaire showed good internal reliability (Cronbach’s α: T1 = 0.88, T2 = 0.94).
2.5 Data-analysis
2.5.1 Participants
Baseline differences between conditions in demographics and mental health were assessed using χ2 tests and ANOVAs. Significant effects of the ANOVAs were examined with pairwise comparisons using Bonferroni adjustment.
2.5.2 Attrition and Missing Data
Baseline differences and attrition between conditions were examined using ANOVAs, and χ2 -tests. The distribution of missingness was examined with Little’s Missing Completely At Random (MCAR) test. All missing data were imputed using multiple imputation (MI). The Markov Chain Monte Carlo (MCMC) imputation procedure was used to conduct MI, and a total of 20 imputed datasets were utilized (Graham et al., 2007). Auxiliary variables used for MI included study variables at T0.
The first hypothesis examined the effectiveness of the intervention compared to the control condition at mid-intervention, post and follow-up measurement. Per measurement moment and per dependent variable, an ANCOVA was performed with condition (BPS, BPAS/BPS and control) as between-subjects factor and baseline levels of the dependent variable as covariate. Pairwise comparisons using Bonferroni adjustment were used to examine significant effects. Pretest-posttest-control group effect sizes (dppc2) were calculated for each measure to analyze the magnitude of change per measurement moment in the experimental conditions compared to the control condition (Morris, 2008). Effect sizes < 0.50 were considered small, between 0.50 and 0.80 medium, and > 0.80 large (Cohen, 1988).
For the second hypothesis, it was first explored whether levels of adherence and engagement differed between conditions and whether they changed over time. ANOVAs were conducted at mid-intervention and post-measurement with condition (BPS, BPAS/BPS and control) as between-subjects factor and dependent variables engagement or adherence. Pairwise comparisons using Bonferroni adjustment were used to examine significant effects. To assess change over time, paired t-test were used. Second, regression analyses were conducted to assess whether adherence and engagement predicted change in the outcome variables. Baseline levels of the dependent variables were added as a covariate to the regression analyses and adherence and engagement at T1 were added as the predictor of the dependent variables at T2. For T3, the same analyses were conducted but with adherence and engagement at T2 as predictors.
The third hypothesis, imagining one’s best self in the past (BPAS) or future (BPS) is more engaging than the control condition which relates to BPS efficacy, was examined using Hayes’(2022) PROCESS macro for SPSS with the default 5.000 bootstrap samples and 95% confidence intervals. A mediation model was tested using model 4. Condition was the multicategorical independent variable, engagement at T1 was the mediator and the dependent variables were assessed at T2. The covariate was baseline levels of the dependent variable. Bootstrap intervals excluding zero indicated a significant mediation effect.
Hayes’ (2022) PROCESS macro was used to test the fourth hypothesis that engagement mediates the association between condition and BPS efficacy, and that these associations are influenced by age. This moderated mediation was tested using Model 8 with condition as the independent multicategorical variable, engagement at T1 as mediator, age as moderator and the dependent variables were assessed at T2. The covariate was baseline levels of the dependent variable. Bootstrap intervals excluding zero indicated significant moderated mediation.
3 Results
3.1 Participants
Figure 2 illustrates participant flow. Of 745 individuals screened, 563 met the eligibility criteria and were invited to complete the baseline measurement. Exclusions were due to high levels of well-being (n = 98) or high levels of anxiety or depression (n = 84). Of those invited, 360 were randomized; non-randomization was due to nonresponse (n = 146), incomplete baseline assessment (n = 17) or failure to download the app (n = 40). Randomized participants were allocated to the BPS (n = 119), BPAS/BPS (n = 123) or control (n = 118) conditions. Not all participants accessed the app (BPS: n = 24; BPAS/BPS: n = 23; control: n = 24) and the final sample consisted of 95 participants in the BPS condition, 100 participants in the BPAS/BPS condition and 95 participants in the control condition.
The mean age of participants was 46.78 years and most participants were female (92.1%), highly educated (73.4%) and Dutch (99.7%). The majority had paid employment (66.6%) and almost half of the participants were married (46.6%). Participants reported levels of well-being that were below average (M = 2.46, SD = 0.61), moderate levels of depressive symptoms (M = 20.31, SD = 6.83) and mild levels of anxiety (M = 6.69, SD = 3.28) (Lamers et al., 2011; Spitzer et al., 2006; Vilagut et al., 2016). There were no significant baseline differences among the three conditions in demographics (p’s range = 0.49–0.93), well-being (F(2, 287) = 0.72, p =.49), or depressive symptoms (F(2, 287) = 1.88, p =.15). Anxiety levels differed between the three conditions (F(2, 287) = 4.65, p =.010). Bonferroni post-hoc analyses showed that levels of anxiety were higher in the BPS condition compared to the BPAS/BPS condition (p =.004) and the control condition (p =.030). Participant characteristics are summarized in Table 1.
3.2 Attrition and Missing Data
The proportion of missing data in the total sample was 0.0% at the baseline measurement, 32.1% at the mid-intervention measurement, 43.8% at post-measurement and 53.1% at follow-up. There were no differences in number of completers versus drop-outs between the three conditions (Mid-intervention measurement: (χ 2 (2) = 0.93, p =.63; BPS: 65.3%; BPAS/BPS: 67.0%; Control: 71.6%; Post-measurement (χ 2 (2) = 0.57, p =.75; BPS: 55.8%; BPAS/BPS: 59.0%; Control: 53.7%; Follow-up (χ 2 (2) = 1.93, p =.38; BPS: 46.3%; BPAS/BPS: 52.0%; Control: 42.1%; see Fig. 2). At the mid-intervention measurement (F(1) = 8.69, p =.003), the post- measurement (F(1) = 17.78, p <.001), and the follow-up (F(1) = 11.68, p =.001), an age difference was found between completers and drop-outs. Completers were relative older than drop-outs (Age completer: T1: M = 47.99; T2: M =.48.97; T3: M = 48.94; Age drop-out: T1: M = 44.22; T2: M = 43.97; T3: M = 44.87). No significant differences were found in the other demographics, well-being, depressive symptoms, or anxiety levels between completers and dropouts across the three measurement points (Mid-intervention measurement: p’s range 0.12 − 0.82; Post-intervention measurement: p’s range 0.19 − 0.92; Follow-up: p’s range 0.10 − 0.98). Little’s MCAR test revealed that data was missing at random (χ²(227) = 217.48, p =.663).
Fig. 2
Participant flow
3.3 Effectiveness of the Intervention
See Table 2 for an overview of the descriptive statistics and pretest-posttest-control group effect sizes. All effect sizes were small and not significant except the effect sizes for the BPS condition compared to the control condition at the post- and follow-up measurement for future life satisfaction and NA.
Table 1
Participants’ characteristics
Total (n = 290) | BPS (n = 95) | BPAS/BPS (n = 100) | Control (n = 95) | ||||||
|---|---|---|---|---|---|---|---|---|---|
Age, years (M, SD) | 46.78 | 10.31 | 46.92 | 9.93 | 46.09 | 10.82 | 47.37 | 10.19 | |
Female gender (n, %) | 267 | 92.1 | 85 | 89.5 | 94 | 94.0 | 88 | 92.6 | |
Dutch nationality (n, %) | 289 | 99.7 | 95 | 100 | 99 | 99.0 | 95 | 100 | |
Educational attainment (n, %) | |||||||||
Low | 3 | 1.0 | 1 | 1.1 | 1 | 1.0 | 1 | 1.1 | |
Medium | 74 | 25.5 | 21 | 22.1 | 27 | 27.0 | 26 | 27.4 | |
High | 213 | 73.4 | 73 | 76.8 | 72 | 72.0 | 68 | 71.6 | |
Marital status (n, %) | |||||||||
Never married | 93 | 32.1 | 29 | 30.5 | 38 | 38.0 | 26 | 27.4 | |
Married | 135 | 46.6 | 47 | 49.5 | 41 | 41.0 | 47 | 49.5 | |
Divorced | 57 | 19.7 | 16 | 16.8 | 20 | 20.0 | 21 | 22.1 | |
Widowed | 5 | 1.7 | 3 | 3.2 | 1 | 1.0 | 1 | 1.1 | |
Well-being (M, SD) | 2.46 | 0.61 | 2.47 | 0.56 | 2.50 | 0.61 | 2.40 | 0.65 | |
Anxiety (M, SD) | 6.69 | 3.28 | 7.49 | 3.36 | 6.13 | 3.22 | 6.47 | 3.12 | |
Depression (M, SD) | 20.31 | 6.83 | 21.37 | 6.70 | 20.06 | 6.57 | 19.51 | 7.15 | |
3.3.1 Mid-intervention
There was a marginally significant effect for NA (F(2, 286) = 2.67, p =.070). Post hoc analyses revealed no significant differences between the conditions. Effects were not significant for optimism (F(2, 286) = 1.50, p =.226), PA (F(2, 286) = 1.25, p =.289), past life satisfaction (F(2, 286) = 0.02, p =.981), current life satisfaction (F(2, 286) = 1.98, p =.140) and future life satisfaction (F(2, 286) = 0.75, p =.475).
3.3.2 Post-intervention
There was a significant effect for future life satisfaction (F(2, 286) = 4.51, p =.012), which was significantly higher in the BPS condition compared to the BPAS/BPS condition (p =.010). For optimism the effect was marginally significant (F(2, 286) = 2.73, p =.067). In the BPS condition levels of optimism were marginally significantly higher compared to the BPAS/BPS condition (p =.065). There were no significant effects for PA (F(2, 286) = 2.15, p =.118), NA (F(2, 286) = 1.57, p =.210), past life satisfaction (F(2, 286) = 2.09, p =.125), and current life satisfaction (F(2, 286) = 0.95, p =.390).
3.3.3 Follow-up
Optimism differed significantly between conditions (F(2, 286) = 3.08, p =.048), with higher levels in the BPS condition compared to the control condition (p =.043). A significant difference was also found for past life satisfaction (F(2, 286) = 5.12, p =.007), with significant higher levels in the BPS condition compared to the BPAS/BPS condition (p =.007) and marginally significant higher levels compared to the control condition (p =.063). Current life satisfaction showed a marginally significant difference between conditions (F(2, 286) = 2.93, p =.055), but post hoc analyses revealed no significant effects between the conditions. There were no significant differences for PA (F(2, 286) = 1.76, p =.173), or NA (F(2, 286) = 1.98, p =.140).
3.4 Adherence and Engagement
See Table 3 for the descriptive statistics of adherence and engagement. Levels of adherence did not differ between the conditions at the mid-intervention measurement (F(2, 287) = 0.17, p =.844) but levels of engagement did (F(2, 287) = 4.31, p =.014), with higher levels in the BPS condition compared to the BPAS/BPS condition (p =.013). At the post-measurement there were no differences between conditions in adherence (F(2, 287) = 0.19, p =.831) or engagement (F(2, 287) = 1.19, p =.306). There was a significant decrease in adherence(t(289) = 9.46, p <.001): during the first week, 84.5% (n = 245) used the app at least once and 14.8% (n = 43) made use of the app every day. During the second week, 70.3% used the app at least once and 7.9% (n = 23) used the app every day. Engagement also significantly decreased from mid-intervention to post-measurement (t(289) = 10.63, p <.001).
See Table 4 for the outcomes of the regression analyses. At T2 and T3, engagement significantly predicted changes in current and future life satisfaction. At T2, engagement predicted changes in almost all outcome variables, except for PA and NA. Adherence was not associated with changes in any of the outcome variables.
Table 2
Descriptives and effect sizes
T0 | T1 | dppc2 | 95% CI | T2 | dppc2 | 95% CI | T3 | dppc2 | 95% CI | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
M | SD | M | SD | M | SD | M | SD | ||||||||||
Optimism | BPS | 3.11 | 0.63 | 3.27 | 0.55 | 0.13 | -0.15, 0.41 | 3.39 | 0.53 | 0.10 | -0.18, 0.39 | 3.39 | 0.49 | 0.25 | -0.03, 0.53 | ||
BPAS/BPS | 3.25 | 0.58 | 3.31 | 0.54 | -0.03 | -0.32, 0.25 | 3.35 | 0.58 | -0.20 | -0.48, 0.08 | 3.39 | 0.49 | 0.00 | -0.28, 0.28 | |||
Control | 3.14 | 0.63 | 3.22 | 0.57 | 3.36 | 0.59 | 3.28 | 0.48 | |||||||||
Past LS | BPS | 3.24 | 1.27 | 3.34 | 1.15 | 0.01 | -0.28, 0.29 | 3.48 | 1.22 | 0.11 | -0.17, 0.40 | 3.67 | 1.20 | 0.18 | -0.10, 0.47 | ||
BPAS/BPS | 2.97 | 1.12 | 3.11 | 1.11 | 0.04 | -0.24, 0.32 | 3.08 | 1.12 | 0.01 | -0.27, 0.29 | 3.16 | 1.00 | -0.02 | -0.30, 0.26 | |||
Control | 3.16 | 1.28 | 3.25 | 1.20 | 3.26 | 1.27 | 3.37 | 1.17 | |||||||||
Current LS | BPS | 3.45 | 1.17 | 3.60 | 1.15 | 0.00 | -0.28, 0.28 | 3.88 | 1.06 | 0.16 | -0.13, 0.44 | 3.90 | 1.09 | 0.29 | 0.01, 0.58 | ||
BPAS/BPS | 3.72 | 1.19 | 3.68 | 1.18 | -0.15 | -0.43, 0.13 | 3.92 | 1.02 | -0.03 | -0.31, 0.25 | 3.78 | 1.11 | -0.03 | -0.31, 0.25 | |||
Control | 3.82 | 1.32 | 3.97 | 1.30 | 4.06 | 1.26 | 3.92 | 1.21 | |||||||||
Future LS | BPS | 3.48 | 0.96 | 3.88 | 0.93 | 0.23 | -0.06, 0.51 | 4.18 | 0.87 | 0.34* | 0.05, 0.63 | 3.99 | 0.91 | 0.39* | 0.10, 0.67 | ||
BPAS/BPS | 3.74 | 0.92 | 3.96 | 0.88 | 0.05 | -0.23, 0.33 | 4.07 | 0.86 | -0.03 | -0.31, 0.25 | 3.93 | 1.07 | 0.06 | -0.22, 0.34 | |||
Control | 3.92 | 1.07 | 4.09 | 1.06 | 4.28 | 1.08 | 4.05 | 0.99 | |||||||||
PA | BPS | 3.42 | 0.53 | 3.41 | 0.53 | 0.14 | -0.14, 0.43 | 3.54 | 0.52 | 0.21 | -0.08, 0.50 | 3.57 | 0.49 | 0.28 | -0.01, 0.57 | ||
BPAS/BPS | 3.42 | 0.60 | 3.32 | 0.64 | -0.02 | -0.29, 0.26 | 3.42 | 0.60 | 0.00 | -0.28, 0.28 | 3.48 | 0.54 | 0.11 | -0.17, 0.39 | |||
Control | 3.54 | 0.63 | 3.45 | 0.55 | 3.54 | 0.55 | 3.54 | 0.48 | |||||||||
NA | BPS | 2.56 | 0.68 | 2.23 | 0.61 | -0.30 | -0.59, -0.02 | 2.01 | 0.54 | -0.32* | -0.60, -0.03 | 2.02 | 0.59 | -0.36* | -0.65, -0.07 | ||
BPAS/BPS | 2.43 | 0.74 | 2.26 | 0.65 | -0.05 | -0.33, 0.23 | 2.03 | 0.52 | -0.07 | -0.34, 0.22 | 2.07 | 0.59 | -0.06 | -0.34, 0.22 | |||
Control | 2.31 | 0.64 | 2.17 | 0.58 | 1.95 | 0.51 | 1.99 | 0.53 | |||||||||
Table 3
Descriptives adherence and engagement
T1 | T2 | ||||
|---|---|---|---|---|---|
M | SD | M | SD | ||
Adherence | Future | 3.73 | 2.57 | 2.64 | 2.36 |
Past/Future | 3.53 | 2.39 | 2.46 | 2.56 | |
Control | 3.67 | 2.37 | 2.46 | 2.19 | |
Engagement | Future | 3.57 | 0.50 | 3.27 | 0.69 |
Past/Future | 3.34 | 0.58 | 3.11 | 0.77 | |
Control | 3.41 | 0.54 | 3.19 | 0.77 | |
3.5 Engagement Mediates BPS Efficacy
See Table 5 for an overview of the results. Engagement at T1 predicted optimism and life satisfaction at T2. Visualizing one’s best possible self in the future was more engaging compared to the control condition, whereas visualizing one’s best possible self in the past was not. Engagement mediated the efficacy of visualizing one’s best possible self in the future compared to the control condition for current and future life satisfaction. No other mediation effects were significant.
Table 4
Adherence and engagement at T1 as predictors of changes in mental health at T2, and at T2 as predictors of changes in mental health at T3
Predictors | Dep. var. | T2 | T3 | ||||
|---|---|---|---|---|---|---|---|
B | SE | p | B | SE | p | ||
Adherence | Optimism | −0.01 | 0.01 | 0.890 | −0.01 | 0.01 | 0.589 |
Engagement | 0.17 | 0.04 | < 0.001 | 0.02 | 0.03 | 0.583 | |
Adherence | Past LS | −0.02 | 0.02 | 0.287 | 0.10 | 0.02 | 0.591 |
Engagement | 0.19 | 0.07 | 0.004 | 0.00 | 0.06 | 0.959 | |
Adherence | Current LS | −0.02 | 0.02 | 0.367 | 0.10 | 0.20 | 0.635 |
Engagement | 0.29 | 0.08 | < 0.001 | 0.15 | 0.07 | 0.002 | |
Adherence | Future LS | −0.01 | 0.02 | 0.356 | 0.00 | 0.02 | 0.981 |
Engagement | 0.31 | 0.07 | < 0.001 | 0.22 | 0.06 | < 0.001 | |
Adherence | NA | −0.01 | 0.01 | 0.185 | −0.00 | 0.01 | 0.852 |
Engagement | −0.01 | 0.04 | 0.777 | −0.02 | 0.04 | 0.600 | |
Adherence | PA | 0.01 | 0.01 | 0.281 | 0.01 | 0.01 | 0.271 |
Engagement | 0.02 | 0.04 | 0.659 | 0.03 | 0.03 | 0.294 | |
3.6 Age, Engagement and BPS Efficacy
See Table 6 for an overview of the results. Age was not associated with optimism, life satisfaction or affect and the effect of the conditions did not interact with age to predict these variables. Age was associated with engagement at T1 but did not interact with the conditions to predict engagement. None of the moderated mediation effects were significant.
Table 5
Mediation analyses with condition as predictor, engagement (T1) as mediator and changes in mental health at T2 as dependent variables
Optimism | Past LS | Current LS | Future LS | NA | PA | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
B | SE | B | SE | B | SE | B | SE | B | SE | B | SE | |
T2 | ||||||||||||
BPS vs. con. | 0.03 | 0.06 | 0.12 | 0.19 | 0.03 | 0.11 | 0.15 | 0.10 | −0.07 | 0.05 | 0.08 | 0.06 |
BPAS vs. con. | −0.07 | 0.06 | −0.01 | 0.09 | −0.05 | 0.10 | −0.07 | 0.08 | 0.01 | 0.05 | −0.04 | 0.06 |
Engagement | 0.15** | 0.04 | 0.17* | 0.07 | 0.27** | 0.08 | 0.27** | 0.07 | −0.00 | 0.04 | 0.01 | 0.04 |
Engagement T1 (Mediator) | ||||||||||||
BPS vs. con. | 0.16* | 0.08 | 0.16 | 0.08* | 0.16* | 0.08 | 0.17* | 0.08 | 0.12 | 0.07 | 0.17* | 0.08 |
BPAS vs. con. | −0.07 | 0.08 | −0.06 | 0.08 | −0.06 | 0.08 | −0.06 | 0.08 | −0.08 | 0.07 | −0.06 | 0.08 |
B | 95% CI | B | 95% CI | B | 95% CI | B | 95% CI | B | 95% CI | B | 95% CI | |
Indirect effect | ||||||||||||
BPS vs. con. | 0.02 | −0.00, 0.06 | 0.03 | −0.00, 0.08 | 0.04* | 0.00, 0.12 | 0.05* | 0.00, 0.11 | −0.00 | −0.01, 0.01 | 0.00 | −0.02, 0.02 |
BPAS vs. con. | −0.01 | −0.04, 0.01 | −0.01 | −0.05, 0.02 | −0.02 | −0.07,0.03 | −0.02 | −0.07,0.02 | 0.00 | −0.01, 0.01 | −0.00 | 0.01, 0.01 |
Table 6
Moderated mediation with condition as predictor, engagement (T1) as mediator, age as moderator, and change in mental health at T2 as dependent variables
Optimism | Past LS | Current LS | Future LS | NA | PA | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
B | SE | B | SE | B | SE | B | SE | B | SE | B | SE | ||
T2 | |||||||||||||
Fut. vs. control | −0.51 | 0.27 | 0.36 | 0.43 | 0.01 | 0.50 | 0.55 | 0.45 | 0.22 | 0.26 | −0.38 | 0.28 | |
Past vs. control | −0.38 | 0.25 | −0.01 | 0.41 | −0.45 | 0.47 | −0.19 | 0.43 | 0.20 | 0.24 | −0.33 | 0.26 | |
Engagement | 0.14** | 0.04 | 0.17* | 0.07 | *0.26 | 0.08 | 0.28** | 0.07 | 0.00 | 0.04 | 0.01 | 0.04 | |
Age | −0.00 | 0.00 | 0.00 | 0.01 | 0.00 | 0.01 | 0.00 | 0.01 | 0.00 | 0.00 | −0.00 | 0.00 | |
Int. age*fut. vs. con | 0.01 | 0.01 | −0.01 | 0.01 | 0.00 | 0.01 | −0.01 | 0.01 | −0.01 | 0.01 | 0.01 | 0.01 | |
Int. age*past vs. con | 0.01 | 0.01 | 0.00 | 0.01 | 0.01 | 0.01 | 0.00 | 0.01 | 0.00 | 0.01 | 0.01 | 0.01 | |
Engagement T1 (Mediator) | |||||||||||||
Fut. vs. control | 0.43 | 0.38 | 0.44 | 0.38 | 0.43 | 0.38 | 0.47 | 0.38 | 0.31 | 0.37 | −0.38 | 0.28 | |
Past vs. control | 0.55 | 0.36 | 0.56 | 0.36 | 0.54 | 0.35 | 0.59 | 0.35 | 0.47 | 0.35 | −0.33 | 0.26 | |
Age | 0.01* | 0.01 | 0.01* | 0.01 | 0.01* | 0.01 | 0.01* | 0.01 | 0.01* | 0.01 | 0.01 | 0.04 | |
Int. age*fut. vs. con | −0.01 | 0.01 | −0.01 | 0.01 | −0.01 | 0.01 | −0.01 | 0.01 | −0.00 | 0.01 | −0.00 | 0.00 | |
Int. age*past vs. con | −0.01 | 0.01 | −0.01 | 0.01 | −0.01 | 0.01 | −0.01 | 0.01 | −0.01 | 0.01 | 0.01 | 0.01 | |
B | 95% CI | B | 95% CI | B | 95% CI | B | 95% CI | B | 95% CI | B | 95% CI | ||
Moderated Mediation | |||||||||||||
Fut. vs. con | −0.00 | −0.00, 0.00 | −0.00 | −0.00, 0.00 | −0.00 | −0.01, 0.00 | −0.00 | −0.01, 0.00 | 0.00 | −0.00, 0.00 | −0.00 | −0.00, 0.00 | |
Past vs. con | −0.00 | −0.01, 0.00 | −0.00 | −0.01, 0.00 | −0.00 | −0.01, 0.00 | −0.00 | −0.01, 0.00 | 0.00 | −0.00, 0.00 | −0.00 | −0.00, 0.00 | |
4 Discussion
The current study examined the specific and generic mechanisms of change in an online BPS intervention using a mobile app. In the course of two weeks, participants were instructed to visualize their best possible self in the future or in the past in different domains such as personal strengths, social relationships and professional achievements. Adherence and engagement with the app were related to mental health changes during the study and engagement and age were related to BPS efficacy.
The first hypothesis was partially supported. Visualizing one’s best possible self in the past did not improve mental health more than the control condition, while visualizing the best possible self in the future had some positive effects. Participants in the BPS condition reported higher future life satisfaction and marginally significant higher levels of optimism post-intervention compared to the BPAS/BPS condition. At follow-up, the BPS condition showed significantly higher levels of optimism and significantly higher past life satisfaction compared to the BPAS/BPS condition. Effect sizes were small whereby the effect sizes in the BPS condition were larger than the BPAS/BPS condition. These findings align with previous research suggesting that thinking about the past is less effective than thinking about the future (Carrillo et al., 2021). When people visualize their BPS through anticipatory emotion regulation (Quoidbach et al., 2015), their belief about mastering future challenges may impact their positive self-concept more strongly than past successes, as these successes can be simply attributed to a “past self” that no longer exists (Schubert et al., 2020). Moreover, visualizing one’s best possible self in the future may be especially effective within a future oriented individualistic Western culture (Sircova et al., 2014). For some individuals, particularly those with adverse past experiences, visualizing the best possible self in the past, may be challenging and counterproductive. A misfit between the person and the activity can backfire positive psychology exercises, as seen in research when expressing gratitude negatively impacted dysphoric individuals (Sin et al., 2011). The found effect sizes for the BPS condition were comparable to previous studies employing online formats and active control conditions (Carrillo et al., 2019). Despite the small effect sizes, the possibility to deliver the BPS app anonymously, at low costs to very large populations suggest strong potential for widespread mental health benefits (Mitchell et al., 2010).
Results supported the second hypothesis: engagement was a stronger predictor of the positive effects of using the app than adherence, regardless whether participants were in the BPS, BPAS/BPS or control condition. After one week intervention use, engagement – but not adherence - predicted changes in optimism and life satisfaction post-intervention. Engagement at post-intervention even predicted changes in current and future life satisfaction four weeks later. These findings support prior research indicating that engagement was a stronger predictor of intervention effectiveness than adherence (Graham et al., 2020, 2021), suggesting that the user’s belief in the app’s usefulness outweigh the actual use of the app (Kelders et al., 2020b). Additionally, the results reflect the dynamic nature of engagement which declined over time. Factors such as novelty, sustained interest, ease of use of the app and the perceived effects of the intervention could be related to this change (Perski et al., 2017).
The third hypothesis was partially supported. Engagement was associated with increases in optimism and life satisfaction, and mediated the beneficial effect of the BPS condition on current and future life satisfaction. No significant effects were found for visualizing one’s best self in the past. Visualizing one’s best possible self in the future may be more engaging than visualizing the best possible self in the past because it was perceived to be more helpful in reaching one’s goal and fostered positive user expectations (Short et al., 2015). Engagement did not predict affect, and this suggest that engagement is stronger related to the cognitive component of wellbeing – namely, life satisfaction – than to the affective component (Lucas et al., 1996). Notably, engagement was a stronger predictor of changes in mental health than the effects of the BPS and BPAS/BPS conditions. This underscores the importance of nonspecific factors, such as engagement, in determining online intervention efficacy and highlights the need for eMental health research to prioritize generic factors.
The results did not support the fourth hypothesis. Visualizing one’s best possible self in the past was not more engaging for older participants, despite expectations of greater personal relevance (Lauriero- Martinez et al., 2017), nor was the BPS condition more engaging for younger participants, as hypothesized (Lyobomirsky & Layous, 2013). However, age was positively related with engagement. Although smartphone use tends to decline with age (Van Deursen et al., 2015), older participants may have found the app to be more helpful and enjoyable than younger users. This finding may help explain the mixed findings in a recent meta-analysis, which reported both positive and negative associations between age and engagement, because engagement was often defined in terms of usage rather than subjective experience (Borghouts et al., 2021). Overall, it can be concluded that the BPAS condition was less engaging and less effective than the BPS condition, including among older participants.
4.1 Limitations and Future Suggestions
Several limitations and suggestions for future research should be noted. First, the self-selected sample consisted primarily of middle-aged, highly educated women, which limits the generalizability of the findings. Individuals with higher education may have more opportunities to pursue their goals, potentially making it easier for them to visualize their BPS (Meevissen et al., 2011). Additionally, women generally hold more positive future expectations than men, which may enhance the benefits of the BPS intervention for female participants (Kooij et al., 2018). Future research should examine whether the efficacy of BPS interventions is predicted by the interaction between temporal frame, gender, and educational level. Second, limiting the sample to individuals with low to moderate levels of well-being may have enhanced treatment effects, as they are more likely to benefit from the BPS intervention. However, we excluded participants with severe symptoms who may have benefitted from the intervention (Tomczyk et al., 2023). Future research should consider including participants across the full spectrum of well-being and symptom severity. In this context, it would be valuable to explore how specific symptom profiles interact with the temporal frame. For example, individuals with depressive symptoms may respond differently to past-oriented BPS visualizations than to future-oriented BPS visualizations (Lefèvre et al., 2019). Third, an elevated cut-off score was used to exclude participants with severe depressive symptoms. However, past research suggest that a cut off score of 16 on the CES-D may overestimate depression prevalence, whereas a higher cut-off score of 34 has greater predictive accuracy (Thomas et al., 2001). Fourth, a substantial number of participants failed to complete questionnaires during or after the intervention. To improve data quality, multiple imputations were used to comply with the intention-to-treat principle. Fifth, follow-up effects of imagining one’s best possible self in the past could not be examined, as participants in this condition were instructed to visualize their best possible future self after one week. It may be interesting for future research to consider participants’ preferred time perspective. Research on Balanced Time Perspective (BTP) - the ability to switch effectively between temporal frames in response to situational and environmental demands – indicates it is beneficial for mental health (Zimbardo & Boyd, 1999). Future research may encourage individuals who typically focus on the future to engage with their past, or vice versa, and this may strengthen BTP and yield beneficial effects. Sixth, the current study was conducted within a Western Educated Industrialized Rich Democratic (WEIRD) context and ignored the cultural origin of positive states, traits, and behaviors (Hendriks et al., 2019). Research suggest that collectivistic cultures are less future oriented than individualist cultures (Sircova et al., 2014) which may help explain why the BPS intervention had less effects on positive affect and well-being among in Chinese adolescents (Wu et al., 2024). Consequently, future research should investigate whether cultural background moderates the effectiveness of different BPS temporal frames. At last, it is important that future research examines how the BPS can be designed to be more engaging, for instance by personalizing app interfaces (Liu et al., 2024).
5 Conclusion
The current study offers an unique contribution to the field by examining both specific and generic factors of a mobile-delivered BPS intervention using a rigorous RCT design. The findings advance the development of eMental health interventions by demonstrating that a BPS intervention aimed at the future can yield small improvement in mental health across age groups. Notably, engagement with the app emerged as a stronger predictor of mental health than adherence or the temporal frame of the BPS exercise. These results underscore the critical role of engagement in the effectiveness of eMental health interventions and highlight its importance for future intervention design.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.