Study population and recruitment

Full details of the study method have been described in the trial protocol. ^{Reference Morgan, Jorm and Mackinnon8} Recruitment to the study was from internet-based sources, including paid advertising with Google, Yahoo! and Facebook, posts on internet forums, promotion in email newsletters and links from other websites. Paid advertising with Google was directed at people who had searched for a short questionnaire online to find out whether they had depression. Other online advertising promoted the study to individuals with mild depressive symptoms as a way of receiving expert information or coping advice about depression via email. Participants joined the study by visiting the website www.moodmemos.com. The Mood Memos website contains clear links to The University of Melbourne and Orygen Youth Health as supporters of the study. Visitors to the Mood Memos website were screened for subthreshold depression with the nine-item Patient Health Questionnaire depression scale (PHQ-9). ^{Reference Kroenke and Spitzer9} Results were categorised as subthreshold depression, an approximation of major depression or no depression. Subthreshold symptoms were defined as two to four symptoms of depression ¹⁰ experienced more than half the days or nearly every day for 2 or more weeks, which have affected work, home or social functioning. Five or more symptoms of depression that had an impact on functioning were categorised as major depression. The no depression category was defined as not falling into the major depression or subthreshold depression categories, i.e. by experiencing fewer than two symptoms or not rating symptoms as having an impact on functioning. Participants were eligible for the trial if they were aged 18 years or over, had subthreshold symptoms of depression (as defined above), were not receiving treatment for depression from a health professional (but maintenance antidepressant medication for 6 months or more was allowed), were a resident of Australia, New Zealand, UK, Ireland, Canada or the USA, and had access to the internet at least weekly. Maintenance medication was allowed so that participants who had experienced a prior episode and were at risk of relapse could take part.

Procedure

The trial was registered at the Australian New Zealand Clinical Trials Registry (ACTRN12609000925246) and was approved by the University of Melbourne Human Research Ethics Committee (HREC 0931313). Individuals who screened positive for subthreshold depression were invited to participate in the study, were assessed for further inclusion criteria and were asked to submit a name and email address. The hyperlink to the pre-intervention assessment was sent to this email address. Once the pre-intervention assessment was completed, participants were randomised to the active group or control group. Immediately following randomisation, participants were sent their first Mood Memos email.

Interventions

The administration of the intervention was fully automated using PHP 5 and a MySQL 5 database. Participants were sent a Mood Memos email twice a week for 6 weeks by the automated system and had no interaction with a therapist. The active group received emails based on self-help strategies endorsed as effective and feasible by depression experts. ^{Reference Morgan and Jorm11} The control group received emails containing general information about depressive disorders (see Appendix for an overview of the content of each Mood Memos email). The design of the active Mood Memos was a crucial component of the study, as merely providing information is not sufficient to change behaviour. The emails employed a variety of techniques to encourage and persuade the recipient to engage in the self-help behaviours. ^{Reference Morgan, Jorm and Mackinnon8} These included persuasive framing, tailoring, goal setting and limiting cognitive load. The emails included an explanation of each strategy, why it would work, tips on how it could be implemented, suggested solutions for implementation barriers, an appeal to commit to implementing the strategy and a reminder about previous strategies. The strategies were ordered in terms of their feasibility to carry out, with the most feasible strategy sent first. Feasibility rankings were taken from the original expert consensus study. ^{Reference Morgan and Jorm11} The control group Mood Memos did not direct any action and were designed not to have any therapeutic value. Their role was to control for the effect of receiving emails with depression-related content.

Outcomes

Assessment points were at baseline, midway through the intervention (3 weeks post-baseline) and at the end of the intervention (6 weeks post-baseline). All assessment was self-rated and undertaken on the Mood Memos website. Participants received up to one reminder email to complete each assessment. The primary outcome was depression symptom severity at post-intervention, assessed with the PHQ-9. ^{Reference Kroenke and Spitzer9} This outcome was also categorised to yield probable depression diagnoses. The PHQ-9 is a well-validated and widely used measure of depression that assesses the frequency over the previous 2 weeks of the nine criterion-A symptoms of a DSM-IV major depressive episode. ¹⁰ The PHQ-9 can be used either as a diagnostic algorithm to make a probable diagnosis of major depressive disorder or as a continuous measure of severity. Scores range from 0 to 27 and cut-points of 5, 10, 15 and 20 represent mild, moderate, moderately severe and severe levels of depression. ^{Reference Kroenke, Spitzer, Williams and Löwe12} Secondary outcomes were more general psychological distress, assessed with the Kessler Psychological Distress Scale (K10), ^{Reference Kessler, Andrews, Colpe, Hiripi, Mroczek and Normand13} and level of functioning, assessed with the Work and Social Adjustment Scale (WSAS). ^{Reference Mundt, Marks, Shear and Greist14} The K10 measures 10 symptoms of mental health in the anxiety–depression spectrum. Scores range from 10 (no distress) to 50 (severe distress) using the Australian scoring method. ^{Reference Andrews and Slade15} It has good internal consistency ^{Reference Donker, van Straten, Marks and Cuijpers16,Reference Kessler, Barker, Colpe, Epstein, Gfroerer and Hiripi17} and is a good discriminator of community ‘cases’ and ‘non-cases’ of DSM-IV disorders. ^{Reference Furukawa, Kessler, Slade and Andrews18} The WSAS measures impairment in work, home management, social activities, private leisure activities and ability to form and maintain relationships. Scores range from 0 to 40, with a score above 20 suggesting moderate to severe psychopathology, scores between 10 and 20 suggest significant functional impairment but less severe symptomatology, and scores below 10 are associated with subclinical populations. The scale has strong psychometric properties, including good internal consistency and good convergence with patient ratings of perceived improvement. ^{Reference Mundt, Marks, Shear and Greist14}

Sample size

The size of the effect of the intervention was anticipated to be small, given that meta-analysis of previous trials of indicated prevention of depression showed a small effect size of d = 0.23. ^{Reference Horowitz and Garber19} Similarly, unsupported internet-based treatment of depression has been shown to have a small effect size of d = 0.25, ^{Reference Andersson and Cuijpers20} consistent with the findings from a meta-analysis published more recently. ^{Reference Cuijpers, Donker, Johansson, Mohr, van Straten and Andersson21} A power analysis indicated that a sample of 393 per condition would give 80% power to detect a small effect size (0.2 standard deviations between conditions) on a continuous outcome measure assuming a correlation of 0.5 between pre- and post-intervention scores. ^{Reference Cohen22} Allowing for a potential drop-out rate of 33%, ^{Reference Christensen, Griffiths and Farrer23} the target total sample size was 1200 participants.

Statistical analysis

Analyses were undertaken on an intention-to-treat basis. Any participants who were randomised but withdrew from the study, or who took up alternative treatments, were included in the analysis as randomised. Primary and secondary outcomes were evaluated using mixed models for repeated measures. Relationships between observations at different measurement occasions were modelled as an unstructured covariance matrix. Degrees of freedom were estimated using Satterthwaite's approximation. Planned contrasts compared changes from pre-intervention to post-intervention between groups. Categorical outcomes were based on the PHQ-9 and were calculated with the same algorithm used during screening, i.e. the categories of subthreshold depression, probable major depression and no depression (or remission from subthreshold depression). Relative risk was calculated and tested for significance. The number needed to treat (NNT) to avoid one person developing major depression and to achieve one remission from subthreshold depression were calculated with 95% confidence intervals (CIs) using the method proposed by Bender. ^{Reference Bender24} A negative confidence limit indicates that the CI for the NNT runs to infinity and includes the possibility that exposure to the intervention may reduce the likelihood of protection or remission. Thus the limit of the CI is the limit of the number needed to treat (harm) (NNTH; see Altman ^{Reference Altman25} ).

Several potential predictors of attrition or depression score post-intervention were explored. These were pre-intervention PHQ-9 score; age; gender (female); history of depression (yes/no); history of bipolar disorder or psychotic disorder (yes/no); mood disturbance (depressed mood or lack of interest/pleasure) at pre-intervention (yes/no); experience of one or more negative life events in the previous 6 months, as assessed by the List of Threatening Experiences (yes/no); ^{Reference Brugha, Bebbington, Tennant and Hurry26,Reference Brugha and Cragg27} confidence in their ability to help themselves with their depressive symptoms (coping self-efficacy), dichotomised to little or no confidence v. moderate or very confident; sought treatment from a health professional for their depressive symptoms during the study (yes/no); and highest level of education (primary/secondary school or trade v. bachelor or postgraduate degree). These predictors were entered into separate logistic regression analyses, with presence of data at post-intervention as the outcome variable. Separate linear regression analyses were run in the active group with post-intervention PHQ-9 score as the outcome variable, adjusting for pre-intervention PHQ-9 score. Predictors that were significant at the P<0.1 level were retained for simultaneous entry in a hierarchical multiple regression analysis with pre-intervention score entered in the first block. Between-group effect sizes (Cohen's d) were calculated by dividing the difference between the two group means at post-intervention by their pooled standard deviation. We tested all effects at the P<0.05 level (two-tailed). Analyses were carried out using SPSS version 18 for Macintosh.

FIG. 1

Recruitment and retention of participants in the Mood Memos study.