Introduction
Smartphone mental health applications (MHapps) have emerged as a vehicle for the delivery of empirically supported technology-based intervention (Economides et al., Reference Economides, Martman, Bell. and Sanderson2018; Firth et al., Reference Firth, Torous, Nicholas, Carney, Rosenbaum and Sarris2017). ‘MoodMission’ (MoodMission Pty Ltd, 2019) is an MHapp directed at anxiety and depression symptoms and wellbeing (Bakker et al., Reference Bakker, Kazantzis, Rickwood and Rickard2018a). Health-promoting activities called ‘Missions’ are provided, using techniques including mindfulness, cognitive restructuring and behavioral activation (BA) (Bakker et al., Reference Bakker, Kazantzis, Rickwood and Rickard2018b).
Engagement with an MHapp is required for outcomes to be achieved (Torous et al., Reference Torous, Nicholas, Larsen, Firth and Christensen2018). Engagement is a broader concept than compliance or adherence and takes into consideration the subjective appraisal of task difficulty and obstacles to task completion (Holdsworth et al., Reference Holdsworth, Bowen, Brown and Howat2014). MHapp engagement can be influenced by privacy concerns (Sagar and Pattanayak, Reference Sagar and Pattanayak2015) and whether users perceive the app as straightforward and useful (Torous et al., Reference Torous, Nicholas, Larsen, Firth and Christensen2018). Previously, engagement with MoodMission was found to predict greater wellbeing (Bakker and Rickard, Reference Bakker and Rickard2019).
Social cognition theories explain engagement with mental health tasks (Kazantzis et al., Reference Kazantzis, L’Abate, Kazantzis, Deane., Ronan. and L’Abate2005), including Bandura’s (Reference Bandura1986) social cognitive theory, where individuals who believe that they are capable of a task are more likely to persevere with the task. Operant conditioning also helps us to consider useful triggers to engagement, where a previously reinforcing experience supports generalization (Skinner, Reference Skinner1938). MHapps that do not offer therapist or peer support can provide informational support to enhance self-efficacy and engagement. Psychoeducation can be provided to promote continued use, intervention satisfaction and perceived helpfulness (Hidalgo-Mazzei et al., Reference Hidalgo-Mazzei, Reinares, Mateu, Nikolova, Bonnin, Samalin, Garcia-Estela, Perez-Sola, Young, Strejilevich, Vieta and Colom2018). However, the type and extent of psychoeducation delivered have not been extensively studied for their impact on engagement in the MHapp context.
MoodMission incorporates BA strategies, promoting participation in positively reinforcing tasks to evoke a sense of mastery and/or pleasure (Martell et al., Reference Martell, Dimidjian, Lewinsohn, Kazantzis, Reinecke and Freeman2010). Mastery-enhancing Missions include making a scrapbook and pleasure-enhancing Missions consist of watching films. Mastery and pleasure have long been discussed in the literature as powerful reinforcers for behaviour change. However, the questions of whether BA should focus on a specific activity type, and whether all activities yield comparable levels of engagement in BA remain unanswered (Dimidjian et al., Reference Dimidjian, Barrera, Martell, Munoz and Lewinsohn2011).
A systematic review concluded that the assessment of the broader construct of engagement is infrequent among published studies (Kazantzis et al., Reference Kazantzis, Brownfield, Mosely, Usatoff and Flighty2017). Additionally, inconsistent assessment methods render comparisons between studies difficult (Ng et al., Reference Ng, Firth, Minen and Torous2019). The Homework Rating Scale – Revised (HRS-II; Kazantzis et al., Reference Kazantzis, L’Abate, Kazantzis, Deane., Ronan. and L’Abate2005) is a measure of engagement and beliefs about the between-session tasks of therapy (i.e. ‘homework’). The HRS-II contains three subscales: engagement, beliefs and consequences (McDonald and Morgan, Reference McDonald and Morgan2013). The scale’s internal consistency is supported by Cronbach’s alphas between .71 and .91, and preliminary analysis (e.g. Sachsenweger et al., Reference Sachsenweger, Fletcher and Clarke2015) has supported its theoretically derived 3-factor structure (Kazantzis et al., Reference Kazantzis, L’Abate, Kazantzis, Deane., Ronan. and L’Abate2005). The HRS-II has been adapted for use in MHapps (i.e. HRS-MA; Bakker and Kazantzis, Reference Bakker, Kazantzis and Kazantzis2020) and was incorporated into the present study following a test of its psychometric properties. Compared with other interventions, MHapps tend to have low retention rates (Bauer et al., Reference Bauer, Glenn, Geddes, Gitlin, Grof, Kessing, Monteith, Faurholt-Jepsen, Severus and Whybrow2020), but as retention is only one small part of engagement, more comprehensive assessment that considers behaviour and cognitive theory determinants of engagement is needed.
Our primary aim was to examine if mode of psychoeducation delivery resulted in different levels of engagement with BA delivered in an MHapp context. It was hypothesized that the extent of engagement, measured by the number of Missions completed and the HRS-MA engagement subscale (HRS-MA-engagement), would be higher for Missions that provided greater support, classified as Missions delivering both written and diagrammatic psychoeducation as opposed to solely written (Hypothesis 1). Second, we sought to compare engagement in mastery- versus pleasure-enhancing BA activities. Past research has not conducted this comparison, although in accordance with self-efficacy theory (Bandura, Reference Bandura1986; and see review by Kazantzis and L’Abate, Reference Kazantzis, Deane, Ronan and L’Abate2005), we hypothesized higher engagement for Missions utilizing mastery- than pleasure-enhancing BA (Hypothesis 2). Our final aim was to investigate the impact of engagement on the user-reported benefit of the app. App appraisal was collected as a post-intervention assessment and therefore was considered as an outcome of engagement. It was hypothesized that engaged participants would report more positive Mission appraisals, measured by the Mission Helpfulness Rating scale (Hypothesis 3a) and the beliefs and consequences subscales of the HRS-MA (HRS-MA-beliefs and HRS-MA-consequences, respectively; Hypothesis 3b). Additionally, participants reporting higher engagement were expected to experience larger mental health gains, assessed by participant distress ratings (Hypothesis 4a) and measures of depression, anxiety and wellbeing (Hypothesis 4b).
Method
Participants
The sample consisted of 238 participants, aged 12 to 63 years (mean = 27.13, SD = 10.84). The demographic frequencies of the sample are presented in Table 1. The majority of participants were within the 20- to 30-year-old age range and female. Most participants had received tertiary education, were not currently studying, and were employed full-time. Participants were members of the general population and were not specifically a clinical sample.
Table 1. Frequencies of participant demographic variables
n = 238. Demographic percentages do not total 100% for age and highest level of education due to non-response.
Participants were voluntarily recruited upon downloading MoodMission to their personal smartphone. App download was free from the iOS and Android App Stores. MoodMission was promoted as a mental health tool on social media sites, with no monetary incentive being provided for app download. Data were collected from app users who used MoodMission since the HRS-MA’s introduction to the app on 29 March 2019. Additionally, participation required app users to provide data until the 30-day evaluation time point. Of the 4260 individuals who used MoodMission within this project’s data collection period, 238 app users satisfied this criterion and were included in the sample. Analyses were performed to compare the demographics of the sample against the demographics of the 4022 MoodMission users who concluded their app use prior to the 30-day time point and are reported as preliminary. There were no other inclusion or exclusion criteria for research participation, apart from requiring access to a smartphone.
Measures
Homework Rating Scale – Mobile Application Version
The Homework Rating Scale – Mobile Application Version (HRS-MA; Bakker and Kazantzis, Reference Bakker, Kazantzis and Kazantzis2020) assessed engagement with and appraisal of MoodMission. Based on an integration of cognitive and behavioural theories (Kazantzis and L’Abate, Reference Kazantzis, Deane, Ronan and L’Abate2005), the measure consists of 12 self-report items rated on a 5-point scale, ranging from 0 (not at all) to 4 (completely/extremely/extensive). Items 1 to 4 belong to the HRS-MA-engagement, items 5 to 9 belong to the HRS-MA-beliefs, and items 10 to 12 belong to the HRS-MA-consequences. The HRS-MA items can be seen in the Supplementary material. The psychometric properties of the HRS-MA had not been investigated prior to the present study. Therefore, a series of analyses on the HRS-MA are reported as preliminary. However, the HRS-II has been evaluated previous to its smartphone adaptation and was found to have strong psychometric properties. The scale has shown high internal consistency, with an alpha of .92 (Hara et al., Reference Hara, Aviram, Constantino, Westra and Antony2017), and a 3-factor structure (Richardson et al., Reference Richardson, Hosemans and Kazantzis2020).
Mission Helpfulness Rating Scale
MoodMission’s Mission Helpfulness Rating scale (MHR) measured how helpful participants found each Mission after completion. The single question ‘how helpful was that?’ is answered on a continuum ranging from 0 (not helpful at all) to 10 (extremely). Individuals slide a pointer along the continuum to identify the point on the scale that quantified their perceived helpfulness. A continuous measure is consistent with prior research to assess perceived helpfulness (Duncan et al., Reference Duncan, Childs and Larson2010).
Subjective Units of Distress Scale
The Subjective Units of Distress Scale (SUDS; Wolpe, Reference Wolpe1969) was used to assess distress, providing a rating between 0 and 10 (Parrish et al., Reference Parrish, Oxhandler, Duron, Swank and Bordnick2016). In MoodMission’s version, a rating of 0 represents ‘not at all’ distressed and 10 represents ‘extremely’ distressed. Similar to the MHR, the rating is given upon sliding a pointer along a continuum, ranging from 0 to 10. The concurrent validity of the scale has been empirically supported (Tanner, Reference Tanner2012).
Patient Health Questionnaire
The 9-item Patient Health Questionnaire (PHQ-9; Kroenke et al., Reference Kroenke, Spitzer and Williams2001) was used to measure depression. Based on the last 2 weeks, items are rated on a 4-point scale ranging from 0 (not at all) to 3 (nearly every day). One item includes ‘little interest or pleasure in doing things’. A total score between 0 and 4 indicates ‘minimal’ depression, 5 to 9 indicates ‘mild’, 10 to 14 indicates ‘moderate’, 15 to 19 indicates ‘moderately severe’ and 20 to 27 indicates ‘severe’. Past evaluation of MoodMission has applied the PHQ-9 to a smartphone context and reported an alpha of .90 (Bakker et al., Reference Bakker, Kazantzis, Rickwood and Rickard2018a). The present study supported the high reliability of the PHQ-9 in a smartphone context, finding an alpha of .88 and McDonald’s omega total of .91.
Generalized Anxiety Disorder Scale
The Generalized Anxiety Disorder scale (GAD-7; Spitzer et al., Reference Spitzer, Kroenke, Williams and Lowe2006) is a 7-item anxiety measure. Individuals reflect on the last 2 weeks and rate the frequency of items, such as ‘trouble relaxing’, between 0 (not at all) and 3 (nearly every day). Mild, moderate and severe anxiety levels are inferred from the cut-off points of 5, 10 and 15, respectively. The GAD-7 previously achieved an alpha of .92 (Bakker et al., Reference Bakker, Kazantzis, Rickwood and Rickard2018a; Spitzer et al., Reference Spitzer, Kroenke, Williams and Lowe2006). The alpha and omega values calculated in the present study were .89 and .92, respectively.
Warwick-Edinburgh Mental Wellbeing Scale
Wellbeing was measured by the Warwick-Edinburgh Mental Wellbeing Scale (WEMWBS; Tennant et al., Reference Tennant, Hiller, Fishwick, Platt, Joseph, Weich, Parkinson, Secker and Stewart-Brown2007). Fourteen statements, including ‘I’ve been feeling useful’, are rated from 1 (none of the time) to 5 (all of the time) based on the last 2 weeks. Total scores between 0 and 32 are classified as ‘very low’, 33 to 40 as ‘below average’, 41 to 59 as ‘average’ and 60 to 70 as ‘above average’. The WEMWBS has high reliability, with an alpha of .92 in a smartphone context (Bakker et al., Reference Bakker, Kazantzis, Rickwood and Rickard2018a). In the present study, the scale achieved an alpha of .88 and omega of .92.
Procedure
A longitudinal design was used to assess participant engagement, appraisal and symptom levels. After downloading MoodMission, participants completed the PHQ-9, GAD-7 and WEMWBS, and provided demographic information. Over a 30-day period, participants decided their frequency of app use. Each app use involved reporting whether they were feeling anxious or depressed, selecting the statement that best matched their mental state, and providing a pre-Mission SUDS rating. Based on the symptom profile reported, the app generated five Mission alternatives from which the participants would self-select one. These Mission alternatives came from a pool of 220 Missions. Participants completed the Mission selected and provided a post-Mission MHR and SUDS rating. After 30 days had elapsed, participants once again completed the PHQ-9, GAD-7 and WEMWBS, in addition to the HRS-MA. All questionnaires were built into the app interface. Data collection from the app was ceased upon completion of the 30-day questionnaires.
Data analysis
Sample size
The target sample size was 266, as estimated for an MHapp intervention group by a multi-level modelling (MLM) study anticipating moderate effects, where power was set at 80% and alpha was set at .05 (Deady et al., Reference Deady, Johnston, Glozier, Milne, Choi, Mackinnon, Mykletun, Calvo, Gayed, Bryant, Christensen and Harvey2018). The present study also anticipated moderate effects, as this effect size has been found for the efficacy of MHapps (Firth et al., Reference Firth, Torous, Nicholas, Carney, Rosenbaum and Sarris2017) and when homework is added to treatments (Kazantzis et al., Reference Kazantzis, Whittington and Dattilio2010). Missions can be considered as similar to homework because they both allow individuals to apply cognitive behavioural therapy (CBT) skills to their everyday environment. A target of 266 participants typically exceeds the sample sizes reported in the current MHapp literature (Bakker and Rickard, Reference Bakker and Rickard2018; Giosan et al., Reference Giosan, Cobeanu, Mogoaşe, Szentagotai, Mureşan and Boian2017).
Preliminary analyses
As there was a large disparity between the number of MoodMission users recruited (N = 4260) and the number included in the sample (n = 238), comparisons were performed between the demographics of the sample and the demographics of the group who stopped using MoodMission prior to the 30-day time point. The Mann–Whitney U-test was conducted to determine whether there was an age difference between the two groups. This test was chosen instead of the independent samples t-test due to normality assumption violation. The ages of the sample (mean rank = 1694.45, n = 237) were significantly lower than of those who stopped using MoodMission (mean rank = 2155.66, n = 4022), U = 373382.50, z = –5.61 (corrected for ties), p < .001, two-tailed. However, this effect was small (r = .10). The Mann–Whitney U-test was also performed to compare the groups for their highest education level. The assumption that group distributions have the same shape was met upon histogram inspection. Highest education level was not found to significantly differ between the sample (mean rank = 1559.69, n = 224) and those who stopped use (mean rank = 1618.05, n = 3003), U = 324171, z = –.95 (corrected for ties), p = .34, two-tailed. Pearson’s chi-square test of independence with an alpha of .05 was conducted to assess whether gender was related to prematurely ending app use. The chi-square test was statistically significant, χ2(2, N = 4260) = 43.51, p < .001, although the effect size was small, with a Cohen’s w of 0.10. A chi-square test investigating whether education status was related to stopping use was statistically significant, χ2(3, N = 4260) = 74.51, p < .001, and the association was small, w = 0.13. Finally, a chi-square test evaluating the independence of employment status and stopping use found a significant relationship between the two variables, χ2(5,N = 4260) = 54.47, p < .001, although this association was small, w = 0.11. All expected frequencies were larger than 5, as assumed for the chi-square test. Therefore, prematurely stopping MoodMission use was related to age, gender, education status and employment status, but not highest education level.
Preliminary analyses were further conducted to test the psychometric properties of the HRS-MA. An alpha of .88 and omega of .93 supported internal consistency. HRS-MA validity was assessed by exploratory factor analysis with varimax rotation. The factor analysis evaluated the number of factors to which the HRS-MA items could be reduced, with comparisons being made to the 3-factor structure previously found for the HRS-II (Richardson et al., Reference Richardson, Hosemans and Kazantzis2020; Sachsenweger et al., Reference Sachsenweger, Fletcher and Clarke2015). These three factors refer to the three subscales: engagement, beliefs and consequences. Inter-item correlations were examined to assess the factorability of the data and were used to conduct a heat map analysis. Items 3 and 4 were excluded due to their low correlations with the other items, all under the recommended minimum threshold of .30 (Jacobs et al., Reference Jacobs, Berduszek, Dijkstra and van der Sluis2017). Upon inspection of eigenvalues and a scree plot, two factors were retained in the factor analysis. Factor analysis revealed that most of the retained items loaded onto both factors. As items 3 and 4 are conceptually important to the construct of engagement, HRS-MA-engagement might not have been represented in the factor analysis. Due to the theoretical support for these engagement items (Kazantzis et al., Reference Kazantzis, L’Abate, Kazantzis, Deane., Ronan. and L’Abate2005), all 12 HRS-MA items were retained in a subsequent factor analysis and the main statistical analyses. When all 12 items were included, the eigenvalues and scree plot supported the retainment of three factors in the factor analysis, which aligns with there being three subscales (see also Supplementary material). Analyses were performed using IBM SPSS Statistics (version 25) and the R Project for Statistical Computing (version 3.5.3).
Data cleaning
Missing values were addressed using the maximum likelihood estimation procedure for the variables assessed via MLM (Dempster et al., Reference Dempster, Laird and Rubin1977). As linear regression is less robust to missing data, a complete sample was used for the variables assessed via linear regression models, with list-wise deletion being performed on cases with missing values (Pepinsky, Reference Pepinsky2018). Screening for outliers was based on inspection of scaled residuals and distance measures, and any outliers were removed (Bakker and Wicherts, Reference Bakker and Wicherts2014).
Assumption checking
Independence and normality assumptions of MLM were assessed (Snijders and Berkhof, Reference Snijders, Berkhof, de Leeuw, Meijer, Goldstein and de Deleeuw2008). Although MLM relaxes the assumption of the independence of observations, as each participant may provide multiple scores, the model’s errors at the highest level of units is assumed to be independent. The normality assumption required the residuals to follow a normal distribution, which was assessed upon Q-Q plot inspection. MLM is perceived as robust to normality violation (Fung and Xu, Reference Fung and Xu2010); however, in the case of significant deviations from normality for data that exhibited behaviour consistent with count data, Poisson MLM was used (Ozonur et al., Reference Ozonur, Akdur and Bayrak2017). Poisson MLM does not assume normality. Linear regression required the independence of observations and normality assumptions to be met, and Q-Q plot inspection supported normality.
Statistical analyses
A two-tailed, paired samples t-test with an alpha of .05 was used to compare the pre- and post-Mission SUDS scores and the baseline and 30-day PHQ-9, GAD-7 and WEMWBS scores. Assumptions of normality and normality of difference scores were met upon Q-Q plot inspection. MLM and linear regression were then used to investigate the predictors and impacts of engagement. MLM is appropriate for hierarchical data, in which participants are nested within a group such as an organization, and repeated measurements, where multiple observations are nested within each participant (Finch, Reference Finch2017). MLM allows variables to vary between and/or within participants. Variables that are held constant are termed ‘fixed effects’ and variables that vary are called ‘random effects’. In the present study, hypotheses comprising variables measured at different time points were assessed via MLM. Two-level MLM was used in which variables under study (level 1) were nested within participants (level 2). Linear regression was considered sufficient for variables collected at a singular time point.
To test Hypotheses 1 and 2, random slope multi-level models A and B were created, using level of support and activity type as predictor variables (level 1). MLM was used to account for participant variability associated with participants completing multiple Missions, as Missions may have provided differing extents of support and some participants may have completed both activity types at different proportions (level 2). Each predictor was dichotomously scored, with the level of support being assigned a value of 1 for Missions accompanied by both a written and diagrammatic description. A value of 0 was given for Missions accompanied solely by a written description (see Supplementary material for examples of Missions differentiated by their level of support). When scoring activity type, mastery-enhancing activities were assigned a value of 1 and pleasure-enhancing activities were assigned a value of 0. Model A used Mission completion as its outcome variable, defined as the total number of Missions completed, indicated by the number of Missions for which a post-Mission rating was completed. Due to Mission completion being a count variable, Poisson MLM was used for Model A. Model B used the HRS-MA-engagement score as its outcome variable.
Random intercept multi-level model C tested Hypothesis 3a, using Mission completion and the HRS-MA-engagement score as predictor variables and the MHR score as the outcome (level 1). Participants may have reported different appraisals as a function of time in addition to engagement, with MLM taking participant variability into consideration (level 2).
Linear regression models were used to test Hypothesis 3b, each using Mission completion and the HRS-MA-engagement score as predictor variables. Model D used the HRS-MA-beliefs score as its outcome variable, whereas Model E used the HRS-MA-consequences score.
Hypothesis 4a was tested using Model F, a random intercept multi-level model which similarly used Mission completion and the HRS-MA-engagement score as predictors. In contrast, the outcome variable used was the change in SUDS score, considered at both the between- and within-participant level (levels 1 and 2, respectively). The change in SUDS score was calculated by subtracting the post-Mission SUDS score from the pre-Mission score, with a positive score indicating an improvement in distress level.
Finally, Hypothesis 4b was tested by linear regression Models G to I. These models used Mission completion and the HRS-MA-engagement score as predictors, with the outcome variable being the change in PHQ-9 score for Model G, GAD-7 score for Model H, and WEMWBS score for Model I. The change in PHQ-9 and GAD-7 scores were calculated by subtracting the 30-day scores from the baseline scores, with a positive score indicating an improvement in symptoms. The change in WEMWBS was calculated by subtracting the baseline scores from the 30-day scores, with a positive score indicating an increase in wellbeing. The analyses were conducted using the R Project for Statistical Computing (version 3.5.3).
Results
Missions completed by each participant ranged from 1 to 65 (mean = 5.46, SD = 8.31, 95% CI [4.37, 6.55]), with the majority of participants completing one Mission and consequently scoring low on this engagement measure. Figure 1 displays Missions completed by the sample according to their level of support and activity type. Completed Missions were typically accompanied solely by text (58.40%) and based on a CBT technique other than BA (72.70%).
Figure 1. Completed missions by their level of support and activity type. ‘Mastery’ denotes missions based on mastery-enhancing behavioural activation (BA); ‘Other’ denotes missions based on a cognitive behaviour therapy technique other than BA; ‘Pleasure’ denotes missions based on pleasure-enhancing BA.
On the HRS-MA, the most common response was in the middle range for the majority of items, with a typical rating of 2 (range 0–4), indicating a moderate level of engagement and appraisal. Conversely, lower scores were typical for the ‘difficulty’ and ‘obstacles’ items of the HRS-MA-engagement and higher scores were generally given for the ‘comprehension’, ‘rationale’ and ‘specificity’ items of the HRS-MA-beliefs. The mean and standard deviation scores of the HRS-MA items are given in the Supplementary material.
The average HRS-MA-engagement score was low to moderate, compared with the maximum possible score of 16. The mean score on the HRS-MA-beliefs, HRS-MA-consequences and MHR were all moderate, with possible score ranges of 0 to 20, 0 to 12, and 0 to 10, respectively. These scores indicate that the sample tended to provide moderately positive appraisals of Missions. Average scores on the pre-Mission SUDS and baseline PHQ-9, GAD-7 and WEMWBS were also located in the middle of these questionnaires’ score ranges, with maximum possible scores of 10, 27, 21 and 70, respectively.
Participants reported a positive change upon completing the post-Mission SUDS and 30-day PHQ-9 and WEMWBS. Participants SUDS significantly decreased with a medium effect size (t 1228 = 24.86, p < .001, d = 0.66). Non-significant differences were observed for the PHQ-9 (t 234 = 1.34, p = .18, d = 0.08), GAD-7 (t 233 = −0.36, p = .72, d = −0.02) and WEMWBS (t 58 = 0.02, p = .99, d = 0.002). According to the cut-off scores of these questionnaires, participants tended to remain exhibiting moderate distress, depression and anxiety severity, and below average mental wellbeing.
Hypothesis 1
The level of informational support was examined for its impact on engagement. When Missions were accompanied solely by a text description, the average engagement level was represented by a fixed effects intercept of 1.24 for the number of Missions completed (SE = 0.13, p < .001) and 6.15 for HRS-MA-engagement (SE = 0.15, p < .001). On average, Missions that provided both a diagram and text description were not significantly associated with an increase in Mission completion (b = 0.02, SE = 0.09, p = .83) or HRS-MA-engagement (b < 0.001, SE < 0.001, p = 1). The random effects slopes revealed minimal individual difference (variance < 0.001).
Hypothesis 2
Engagement with mastery-enhancing BA activities was compared to engagement with pleasure-enhancing BA activities. Pleasure-enhancing activities were associated with a fixed effects intercept of 1.60 for Mission completion (SE = 0.20, p < .001) and 6.31 for HRS-MA-engagement (SE = 0.29, p < .001). Mastery-enhancing activities did not significantly predict an increase in Mission completion (b = 0.004, SE = 0.18, p = .98) or HRS-MA-engagement (b < 0.001, SE < 0.001, p = 1). The random effects slopes revealed minimal individual difference (variance < 0.001).
Hypothesis 3a
Engagement was assessed for its impact on the MHR score. An increase in Mission completion did not significantly predict an increase in the MHR score (b = 0.02, SE = 0.02, p = .31). Similarly, the HRS-MA-engagement score failed to significantly predict an increase in the score on the MHR (b < 0.001, SE = 0.13, p = 1).
Hypothesis 3b
The effect of engagement on scores on the HRS-MA subscales assessing Mission appraisal, HRS-MA-beliefs and HRS-MA-consequences, was considered. Mission completion and HRS-MA-engagement accounted for a significant 38% of the variability in HRS-MA-beliefs, R 2 = .38, adjusted R 2 = .35, F 2,45 = 13.87, p < .001. Higher scores on Mission completion and HRS-MA-engagement were significantly associated with an increase in HRS-MA-beliefs, b = 0.16, SE = 0.05, p = .003 and b = 0.88, SE = 0.21, p < .001, respectively. A significant 59% of the variability in HRS-MA-consequences was attributable to the two engagement measures, R 2 = .59, adjusted R 2 = .57, F 2,45 = 31.73, p < .001. Mission completion and HRS-MA-engagement each significantly predicted an increase in HRS-MA-consequences, b = 0.08, SE = 0.03, p = .003 and b = 0.74, SE = 0.10, p < .001, respectively.
Hypothesis 4a
The present study investigated the impact of engagement on the change in SUDS score. The average change in SUDS score decreased for each Mission completed (b = –0.001, SE = 0.03), although this relationship was not statistically significant (p = .98). Likewise, an increase in HRS-MA-engagement did not significantly predict a decrease in the average change in SUDS score (b = –0.16, SE = 0.21, p = .47).
Hypothesis 4b
Engagement was assessed for the ability to predict a change in PHQ-9, GAD-7 and WEMWBS score. Mission completion and HRS-MA-engagement contributed to a non-significant 2% of the variability in the change in PHQ-9 score (R 2 = .02, adjusted R 2 = –.03, F 2,45 = 0.34, p = .71), 1% of the variability in the change in GAD-7 score (R 2 = .01, adjusted R 2 = –.03, F 2,45 = 0.23, p = .77), and 15% of the variability in the change in WEMWBS score (R 2 = .15, adjusted R 2 = –.04, F 2,9 = 0.80, p = .48). An increase in the number of Missions completed was not significantly associated with an increase in the change in PHQ-9 score (b = 0.06, SE = 0.08, p = .45), GAD-7 score (b = 0.02, SE = 0.07, p = .79) or WEMWBS score (b = 0.03, SE = 0.59, p = .96). Similarly, an increase in HRS-MA-engagement did not significantly predict a change in PHQ-9 score (b = –0.12, SE = 0.33, p = .72), GAD-7 score (b = –0.20, SE = 0.29, p = .50) or WEMWBS score (b = 0.99, SE = 0.80, p = .25).
Discussion
The present study examined engagement with smartphone-delivered CBT interventions in the MoodMission app. Engagement was measured by the number of Missions completed and the HRS-MA-engagement. We tested the hypothesis that interventions with diagrammatic and text descriptions would have higher levels of engagement than those with text alone (Hypothesis 1). We also tested the hypothesis that the type of BA intervention (i.e. mastery enhancing vs pleasure enhancing) would yield different engagement levels (Hypothesis 2). Significant differences were not obtained and Hypotheses 1 and 2 were not supported. One possible reason for our results is the limited amount of psychoeducation delivered per technique in MoodMission, a natural limitation of the amount of information that can be presented on a small screen. Diagrams were also limited in their complexity and range. However, this is the first study to examine the mode of rationale delivery and the type of CBT intervention within a smartphone app and further research is required. Future studies should investigate the effects of other levels of psychoeducation support delivered via MHapp, including videos, photos, vignettes and audio recordings (Hidalgo-Mazzei et al., Reference Hidalgo-Mazzei, Reinares, Mateu, Nikolova, Bonnin, Samalin, Garcia-Estela, Perez-Sola, Young, Strejilevich, Vieta and Colom2018).
We did find significant relationships between measures of engagement and HRS-MA beliefs and consequences. The HRS measure has been adapted from the face-to-face delivery of CBT and examines both engagement and the theoretically meaningful determinants of engagement (see review in Kazantzis et al., Reference Kazantzis, L’Abate, Kazantzis, Deane., Ronan. and L’Abate2005). We tested the HRS modification for MHapps (HRS-MA: Bakker and Kazantzis, Reference Bakker, Kazantzis and Kazantzis2020) and found acceptable levels of reliability and factorial validity. Therefore, it is meaningful that participants who were more engaged with the app and used it more frequently also had a more positive belief in their app use, and experienced a greater sense of collaboration, progress and mastery, and having a clearer rationale for participating in the activities. Our findings underscore the importance of theory as the basis for designing the manner in which rationales for interventions are communicated to MHapp users (Torous et al., Reference Torous, Nicholas, Larsen, Firth and Christensen2018). Future research should examine whether engagement–outcome relations in smartphone-delivered CBT interventions are similar to those found in face-to-face therapy (Kazantzis et al., Reference Kazantzis, Luong, Usatoff, Impala, Yew and Hofmann2018).
The present study had limitations that should be acknowledged. First, the present investigation was a naturalistic study; there was no experimental control over which Missions were chosen by participants and for what reasons. This offers advantages, such as improved ecological validity, but may have introduced a degree of variance that made it difficult to detect meaningful effects. The present sample was also modest compared with previous research with MoodMission (e.g. Bakker and Rickard, Reference Bakker and Rickard2019) and there was a high level of attrition. Larger sample sizes are required to detect the small effect sizes of MHapp interventions, particularly when those participants are naturally recruited as typical app users without financial incentive. Retention rates for MHapps without financial incentive are commonly low when compared with other interventions (Bauer et al., Reference Bauer, Glenn, Geddes, Gitlin, Grof, Kessing, Monteith, Faurholt-Jepsen, Severus and Whybrow2020), so MoodMission’s retention rate (5.5%) was consistent with average Apple and Android app retention rates in the general population; the rate was 4% after 90 days (Statista, 2016).
The present study provides further support for the HRS items, and the modification as a measure of engagement in an MHapp context. Other MHapps could incorporate the HRS-MA and assess for user engagement. Further research is required to examine the engagement–outcome relationship, its mediators, and those stable individual characteristics that serve as moderators, using a larger sample size. Increasing understanding of engagement, including how it can be measured and the importance of user appraisal, can be used as a basis for enhancing MHapp design.
Acknowledgements
None.
Financial support
None.
Conflicts of interest
None.
Ethics statements
The authors abided by the Ethical Principles of Psychologists and Code of Conduct as indicated by the ARC/NHMRC guidelines in accordance with the APS. The research was approved by the Monash University Human Research Ethics Committee (reference number 18294). MoodMission users consented to their data being used for research by consenting to MoodMission’s privacy policy. For minors, the privacy policy contains an explanatory statement to be read by guardians.
Supplementary material
To view supplementary material for this article, please visit: https://doi.org/10.1017/S1352465820000922
Comments
No Comments have been published for this article.