Eligibility criteria

Participants were required to meet the following criteria: Australian resident aged 18 to 75 years; own an internet-enabled mobile phone; have access to a desk-top computer with internet capability; have a valid email address; and report symptoms of mild-to-moderate depression, anxiety and/or stress, defined as a total score of 27–63 inclusive on the Depression Anxiety and Stress Scales (DASS; [17]). Consistent with our public health focus on mild-to-moderate symptoms, people were excluded if they scored 64 or more on the DASS (severe symptomatology), answered positively to questions asking about suicidal thoughts, intent and/or previous suicide attempts, or met criteria for psychotic symptoms, as measured by the Psychosis Screening Questionnaire (PSQ; [18]).

Sample & setting

Based on data from our proof of concept study [14], our targeting of mild-to-moderate symptoms, and the fact that myCompass is a self-help intervention without therapist support, we estimated a between-group effect size of d = 0.2. Power calculations established that for 80% power and an alpha of .05, 400 people per group would be required to detect a difference if one was present. To allow for sample attrition, we set a sample size target of n = 650 per group. Participants were recruited nationally and online between October 2011 and March 2012 via social media (Facebook and Twitter), websites of the Black Dog Institute (BDI), corporate and government organisations, advertisements placed on national radio and in print media, and the BDI volunteer research register. Due to time and funding restrictions, it was not possible to extend recruitment activities beyond this period. Written informed consent was provided by all participants. The RCT was approved by the Human Research Ethics Committee at the University of New South Wales, Sydney, Australia (HREC 100019), and registered as Australian New Zealand Clinical Trials Registry ACTRN 12610000625077.

Design

A mixed factorial repeated measures design was employed, with full randomisation to three conditions.

Randomisation

A research assistant not involved in the RCT randomised participants after baseline using computerised random numbers. Allocation was either to the myCompass, AC or WL condition. Participants received advice of their group assignment by email.

Interventions

myCompass is a fully-automated, self-help, public health intervention, that is tailored to the user and has no therapist input. Real-time self-monitoring of symptoms (e.g., problem moods, thoughts and behaviours) via mobile phone and/or computer is a key therapeutic feature. Users can self-monitor three symptoms of their choice at any one time, selected from a list of 20, or three that are recommended to them by the program (e.g., confidence, worry, irritability, motivation, diet, and medication use). Each symptom is rated on a 10-point scale (e.g., “how confident do you feel right now”, “how worried do you feel right now”, “how satisfied are you that you have taken your prescribed medication today”). At the time of rating, users also provide contextual information about where they are, what they are doing and who they are with, using a series of drop-down menus. Users can schedule short message service (SMS) or email reminders to facilitate self-monitoring (frequency of reminders determined by the user); receive and print graphical feedback about their monitoring, including contextual information, on their phone or computer (to monitor change and assist identification of triggers); and elect to receive helpful facts, mental health-care tips or motivational statements by SMS or email.

Evidence-based and interactive psychological modules that users can complete via the internet on their computers are another key element of myCompass. The program contains 12 skill-building modules derived from CBT, Interpersonal Psychotherapy, Problem-solving Therapy and Positive Psychology that cover topics such as Managing Fear and Anxiety, Tackling Unhelpful Thinking, Managing Loss and Major Life Change, and Solving Problems. Each module comprises three 10-minute sessions and includes activities for users to complete on the computer. Modules also recommend home practice tasks for completion between the weekly sessions to promote skill generalisation. Users can complete the modules of their choice, or those recommended to them by the myCompass program.

Tailoring of the myCompass program self-monitoring and module recommendations occurs via users’ responses to a profile questionnaire completed at registration. This questionnaire asks users to rate the extent to which they experience a range of symptoms of depression, anxiety, and stress (e.g., worry, irritability, difficulties with motivation and concentration). Targeting the three highest rated symptoms, in-built algorithms generate personalised feedback to the user about the self-monitoring dimensions and psychological modules that may be of greatest benefit (i.e., each dimension and module is weighted according to its clinical relevance for each symptom, and those with the highest weightings are recommended).

User privacy is managed by a password protected log-on, and by ensuring that user generated data (i.e., self-monitoring ratings) are not stored on the users’ phone but instead transferred via the internet using secure sockets layer protocols (which encrypt transmitted data by rendering it unreadable to anyone other than the intended recipient) and by storing the data in secure servers. Registering to use myCompass is free and users are not billed for the SMSs they receive.

Participants randomised to the myCompass intervention were emailed instructions for accessing and logging onto the program. They were advised that they had full access to the program for seven weeks, and were encouraged to use the program ad libitum during this time. However, it was recommended that they complete a minimum of two modules and monitor daily at least three moods or behaviours during the intervention phase. Access to the program was withdrawn at the end of seven weeks.

Attention control participants received a control mental health program matched to the active intervention on duration and mode of delivery. Each week for seven weeks, they received a fact sheet containing information about depression, anxiety or stress sent to their email address. The information was designed to be read on computer in approximately 10 minutes, and to be credible but void of management advice or treatment strategies. They also received on their mobile phones weekly SMS messages containing brief factual statements about depression, anxiety and stress. The mobile phone statements were also therapeutically inactive, but chosen to ensure that the control program had face validity. Messages varied in length from 160 to 275 characters so as to match, as far as possible, the minimum amount of SMS content that myCompass participants received each week, namely one self-monitoring reminder (160 characters) plus an optional motivational statement, mental health care tip or fact (average length 95 characters).

Waitlist participants did not receive emails or SMSs during the intervention phase, but received full access to the myCompass program at the end of the seven weeks.

By including both the AC group and the WL group, we were able to control for non-specific effects of the intervention. We were also able to test whether any effects seen during the intervention phase were replicated when the WL control group subsequently completed the myCompass program.

Procedure

Data collection took place between October 2011 and October 2012. All study consent, screening and questionnaire data were collected online using online survey software. Individuals not meeting screening criteria received automated feedback explaining the reason/s for their ineligibility, and referring them to other online resources of the BDI. Individuals with symptoms in the severe range were also advised to seek face-to-face support from a health professional and were provided the contact details of crisis support services.

Eligible participants completed a baseline questionnaire prior to randomisation, a post-intervention questionnaire administered at eight weeks, and a follow-up questionnaire administered 12 weeks later for participants in the myCompass and AC groups, and 19 weeks later for the WL group. At each assessment point, participants accessed the appropriate study questionnaire via a link sent to them in an email message.

Primary outcome

The Depression, Anxiety and Stress Scales measure (DASS-21; [17]) is a widely used self-report measure of depression, anxiety and stress with high internal consistency, acceptable test-retest reliability [19], and yields reliable and valid data when used in an online format [20]. Respondents are asked to indicate the frequency with which they experienced symptoms of depression, anxiety and stress over the previous week. Total scores range from 0 to 126 and subscale scores range from 0 to 42, with higher scores indicating greater symptom severity.

Secondary outcomes

The Work and Social Adjustment Scale (WSAS; [21]) assesses the degree to which mental health problems interfere with day-to-day functioning in five domains: work, social leisure activities, private leisure activities, home-management, and personal relationships [21]. The measure provides an assessment of the experiential impact of mental health symptoms from the sufferer’s point of view. Scores range from 0 to 40, with higher scores indicating poorer adjustment. Meyer et al. [22] provide data supporting the psychometric adequacy of the WSAS when administered in an online format.

Demographic and clinical information collected within the baseline questionnaire included age, gender, marital status, educational level, employment status, frequency of mobile phone and internet use, frequency of depressive and anxiety episodes, and age of first episode. At each assessment point, health service utilisation for mental health reasons (e.g., visits to general practitioner, mental health professionals and alternative health practitioners) and use of antidepressant and anxiolytic medication were also assessed. Satisfaction with the program was measured at post-intervention for participants in the myCompass and AC groups using a study-specific measure assessing users’ perceptions of program usability, content, flexibility and functionality.

Adherence, defined as the extent to which participants engaged with the intervention [23], was examined for the myCompass group with respect to three indices, namely, frequency of logins, frequency of self-monitoring, and number of modules attempted.

Statistical analyses

Statistical analyses were completed with SPSS 20.0 software. Characteristics of the three groups at baseline were compared using chi-square tests for categorical variables and analysis of variance (ANOVA) for continuous variables. Chi-square and ANOVA were also used to compare baseline characteristics of participants who did (‘non-dropouts’) and did not (‘dropouts’) return completed questionnaires at post-intervention and follow-up, to explore possible biases in attrition.

Effects of the myCompass intervention on study outcomes were evaluated using intention-to-treat (ITT) analyses that included data from all participants who completed the baseline assessment and any follow-up assessment. Strategies for dealing with missing data in longitudinal studies vary, so we adopted two recommended techniques for analysing incomplete datasets, namely, mixed models repeated measures (MMRM) and multiple imputation (MI; for a review see [24]). The two analyses yielded similar outcomes and, as MMRM is the dominant strategy adopted in studies of web-based interventions, the findings derived from MMRM are presented here.

In MMRM, no participant is removed from the analysis because all available data is used to obtain parameter estimates. In the present study, restricted maximum likelihood (REML) was used to estimate model parameters and error degrees of freedom were calculated using Satterthwaite’s approximation [25]. Analyses assumed a compound symmetric structure, in line with Fairclough’s recommendation that the covariance structure be restricted in situations where attrition is high [26].

Post-intervention analyses involved a series of 3 groups (myCompass, AC and WL) by 2 times (pre-intervention, post-intervention) repeated measures models. Because WL participants were exposed to the myCompass intervention during the follow-up period, repeated measures models analysing follow-up data comprised 2 groups (myCompass and AC) by 3 times (pre, post and follow-up). For each analysis, scores at baseline on the outcome variable of interest was entered as a covariate. Significant group by time interaction effects were examined using sets of Bonferroni adjusted interaction contrasts to compare between-group differences in mean change on the outcome measure from baseline to post-intervention (i.e., intervention period) and from post-intervention to follow-up (i.e., follow-up phase), as appropriate. All effects were tested at the p < 0.05 level, with adjustment according to the number of contrasts in each set.

Cohen’s d was calculated following the procedure outlined in Ruwaard et al. [27] and using both observed and estimated marginal means. For all outcome measures, within-and between-group differences were standardised to Cohen’s d using the pooled standard deviation of the observed scores obtained at baseline.

To determine the replicability of treatment effects resulting from the intervention, primary and secondary outcome data from WL participants who used the myCompass program at the end of their waiting period were examined at post-intervention and follow-up using paired samples t-tests and Cohen’s d.

Randomisation and study attrition

Details of enrolment into the trial, organised according to the CONSORT guidelines [28], are shown in Figure 1. Of the 3561 people who expressed an interest in the study, 720 were randomised. Among those who returned completed screening information, reasons for ineligibility included: symptoms too severe (975, 73.1%); no internet-enabled mobile phone (167, 12.3%); symptoms in the normal range (150, 11%); no access to the internet or mobile phone (49, 3.6%) and not an Australian resident (17, 0.01%). Fifteen people subsequently withdrew from the study and, in the absence of consent to use their data, we excluded them from the analyses.

The rate of attrition for the total sample was 27.9% at post-intervention and 51.4% at the 3-month follow-up assessment (see Figure 1). At post-intervention, attrition was higher among employed (27.9%) than unemployed participants (18.4%; X ² ₁ = 4.04, P = 0.04), slightly higher in males (31.6%) than females (24.4%; X ² ₁ = 3.77, P = 0 .05) and higher in the myCompass group (45.5%) than the AC (20.7%) and WL groups (13.2%; X ² ₂ = 67.84, P < 0.001). Those who did not complete the post-intervention assessment also reported significantly lower DASS Anxiety scores at baseline (t ₇₀₃ = -1.96, P = 0.05). Attrition rates at 3-month follow-up were 56.7% and 32.1%, for the myCompass and AC groups respectively, and differed significantly (X ² ₁ = 29.25, P < 0.001). Demographic variables did not differ between dropouts and non-dropouts at follow-up, however, dropouts reported significantly less functional impairment at baseline on the WSAS (t ₇₀₃ = -2.01, P = 0.04).

For the myCompass group, there were no significant baseline differences between non-dropouts and dropouts at post-intervention and follow-up on any of the demographic and clinical history variables.

Outcomes at post-intervention

Outcomes at follow-up

Figures 2 and 3 depict the estimated marginal means for the myCompass and AC conditions on the primary and secondary outcome measures at baseline, post-intervention and follow-up.

Effect size

Waitlist group results

WL participants were given access to myCompass at the end of the waiting period, and 52% (103) returned a questionnaire after using the myCompass intervention for seven weeks. Paired samples t-tests showed significant immediate improvement for participants on the three DASS subscales and the total DASS (t ₁₀₁ range = 1.97-3.10; P range 0.001 to 0.05) and on the WSAS (t ₉₅ = 2.94; P = .004). Follow-up data were provided by 66% (68) of this group and showed that these improvements were maintained at follow-up (t ₆₇ range = -1.22-1.30; NS). Within-group effect sizes calculated immediately after the intervention and using observed means were small (d range = .17 for DASS Anxiety to .31 for DASS Depression).

Program adherence and satisfaction

Participants in the intervention group logged in to use the myCompass program an average of 14.7 times (s.d.=16.7; range 0 to 105) and self-monitored an average of 49 times (s.d.=54.1; range 0 to 262) during the 7-week intervention period. The mean number of modules completed was 1.6 (s.d.=1.7; range 0 to 9). There was no difference at baseline between people who did and did not log in to use myCompass during the intervention period. None of the adherence indices correlated with demographic, clinical history and primary and secondary outcome data obtained at baseline.

Scores on the program satisfaction measure ranged from 1 (dissatisfied) to 5 (extremely satisfied). On average, ratings of program satisfaction were above the mid-points of the scale and did not differ for participants in the myCompass (mean rating = 3.86, s.d.=0.82) and AC (mean rating = 3.66, s.d.=0.85) conditions (t ₂₉₉ = 1.92; NS). Eighty-three per cent of myCompass participants reported that they would recommend the program to others, and 87.4% indicated that they would happily use the program again.

Limitations

Limitations of this research need to be acknowledged. In common with previous internet trials [23, 33], dropout attrition was high, especially for the myCompass group, and rates of engagement for myCompass participants with the program content were highly variable (and in some instances minimal). Inspection of possible biases due to attrition showed that dropouts were more likely to be male and employed, thus reducing our confidence in generalising to these groups. While our statistical methods accounted for dropout attrition and non-completion, we cannot discount the possibility that dropouts from the intervention group were less satisfied with the program and/or experienced less positive outcomes. Further research is required to examine the predictors of usage of the myCompass program, and relations between program usage and symptom and functional gains.

To the best of our knowledge, this is the first study to demonstrate the effectiveness of mobile phone and computer technology to deliver a fully automated, self-help intervention for a community sample with depression, anxiety and stress. Future research is needed to isolate the relative contributions of the mobile phone (e.g., self-monitoring, SMS messages) and computer-based (e.g., psycho-educational modules) elements of the intervention. We are currently planning a dismantling study comparing different program elements (e.g., mobile versus email self-monitoring; modules only versus modules and self-monitoring), in order to identify the ‘active’ program elements. Further research to quantify work productivity improvements and monetary savings, and to investigate the short-and long-term effectiveness of myCompass in more severely distressed individuals is also warranted.