Encouraging results of online intervention for problems including panic disorder, anxiety, posttraumatic stress [14-17], and depression have been reported [14,18,19] and such treatments can have similar efficacy to face-to-face therapy [19-23]. Reviews of the evidence report that although self-guided interventions for depression have benefit to users [14,18,19], even low-intensity guided support from coaches or therapists helps to increase adherence to online mental health treatments [24,25]. Typically, therapeutic effects can be achieved by online interventions that offer low-intensity guidance of less than 3 contact hours in a 6-week program. Guided support can also provide a secondary “safety net” for individuals whose symptoms deteriorate during online treatment.

We developed a PND intervention (MumMoodBooster) with low-intensity guided support based on cognitive behavioral therapy (CBT), which is an established treatment of choice for depressive disorders with its efficacy supported by much research [26,27]. MumMoodBooster was adapted from our Getting Ahead of Postnatal Depression program, which is specifically adapted for the needs of postnatal women (eg, presenting behavioral strategies before cognitive content. Postnatally depressed women, overwhelmed by the demands of infant care, are often not ready to engage in cognitive tasks prior to some behavior change [28]). Previously, we have reported fully on the formative development and systematic usability testing of the MumMoodBooster program [11]. The program developmental process followed an iterative staged approach [29] recommended for development and testing of behavioral interventions [30,31]. We have also demonstrated clinical efficacy in an uncontrolled feasibility trial [13], which showed excellent adherence and acceptability. Of all users, 87% completed all six program sessions; of those women meeting diagnostic criteria for depression at baseline, 90% no longer met criteria after treatment [13].

MumMoodBooster is now one of two evaluated Internet interventions for PND. O’Mahen and colleagues have also developed and tested Netmums, a guided online behavioral activation (BA) treatment [12,32]. MumMoodBooster is a briefer intervention, (6 sessions vs 12 in Netmums) and includes cognitive therapy as well as BA, but both programs include regular telephone support. This is accomplished by a low-intensity, nontherapeutic coaching role in the case of MumMoodBooster and more frequent calls by mental health workers trained in a “high-intensity perinatal-specific BA approach” in the case of Netmums. In the antenatal period, a recently pilot-tested computerized CBT intervention for depression among pregnant women has also shown significant improvements on self-report psychometric measures [33].

Here we report on the efficacy of the MumMoodBooster intervention delivered with low-intensity guided telephone support in a randomized controlled trial (RCT) compared to a treatment as usual (TAU) condition. We included a sample of women diagnosed with a depressive disorder and reassessed diagnostic status 12 weeks after enrollment. We hypothesized that MumMoodBooster would lead to a reduction in depressive symptomatology and an increased rate of remission from the diagnosed depressive episode compared to TAU. Current evidence points to similar potential for efficacy between online and face-to-face CBT for depression in general. Our study sought first to establish whether a PND-specific online CBT program is clinically effective compared to TAU care practices.

This was a parallel 2-group RCT comparing MumMoodBooster to TAU. The main outcomes were remission from the depressive episode and severity of symptoms of depression at 12 weeks postenrollment. See Multimedia Appendix 1 for the study's CONSORT-EHEALTH checklist [68].

The study was approved by the Austin Health Human Research Ethics Committee (approval number H2012/04682) and informed consent was obtained from all participants.

Recruitment occurred between March 2013 and July 2014. Participants were women resident in Australia aged 18 years and older with a child aged less than 1 year. A stepped process was used to recruit eligible participants: initial screening, clinical assessment, baseline data acquisition, and randomization. The research project was available to women across Australia, in both rural and metropolitan areas. Marketing focused on Internet campaigns using Google AdWords, Facebook, and Twitter. The project was also advertised to Maternal and Child Health Centres in Melbourne who were encouraged to direct clients to the project. Messages prompted interested individuals to obtain more information and to begin a screening process by visiting the secure project recruitment website or to call project staff directly.

Screening criteria were determined online as follows: Australian residency, 18 years of age or older, English speaking, less than 1 year postpartum, Internet access with regular email use, an Edinburgh Postnatal Depression Scale (EPDS) [34,35] score of 11 to 23, no current treatment for depression (medication or psychotherapy), and a score of less than 3 on item #10 of the EPDS (indicating frequent thoughts of self-harm).

Individuals who satisfied screening criteria were emailed a participant information and consent form to complete and return by email. Consenting women were subsequently telephoned by a psychologist/psychology trainee to explain the study and to schedule a clinical assessment. At this stage, women who scored 1 to 2 on item #10 of the EPDS were asked a series of questions following the risk assessment protocol of Simon and colleagues [36] to determine intent, lethality, access to means, and history of suicide attempts. Those deemed to be at risk for suicide were excluded and referred to receive immediate crisis attention.

Following initial telephone contact, eligible consenting women were assessed by telephone by a clinical psychologist/psychology trainee using the Structured Clinical Interview for DSM-IV (SCID-IV) [37,38]. Inclusion criteria based on the SCID-IV assessment were (1) meeting criteria for a major depressive disorder or (2) meeting criteria for a minor depressive disorder. Exclusion criteria were (1) current substance abuse, (2) current and past manic/hypomanic symptoms, (3) posttraumatic stress disorder, (4) alcohol abuse or dependence, (5) depression with psychotic features, (6) risk of suicide as per risk protocol, and (7) current active treatment for depression (medication or psychotherapy). All assessed women were asked to nominate a contact health professional (eg, a general practitioner) to whom the project could send notification of her diagnosis.

Women satisfying all eligibility criteria were told verbally about the unique log-in they could use to complete the baseline (prerandomization) questionnaires.

Women who completed the baseline assessment questionnaires were randomized immediately online to either MumMoodBooster or to TAU. The randomization procedure used a 1:1 allocation ratio and a pregenerated permuted blocks allocation schedule with the sequence concealed from the researchers consistent with CONSORT standards [39]. Treatment allocation to condition was revealed in a phone call. Women allocated to MumMoodBooster were told verbally about how to begin accessing the program. Given the nature of the intervention, participants could not be blinded to treatment beyond the point of allocation.

Safety monitoring was used to provide crisis intervention and referrals to mental health specialty care if needed and a reminder of the emergency contacts provided at baseline assessment. For women in the MumMoodBooster condition, these were carried out by the allocated telephone coach. For women in TAU, a separate safety monitor was allocated (from the same pool of staff as the telephone coaches). All study participants were monitored via safety calls on five occasions: at baseline and at weeks 3, 5, and 9 (immediately posttreatment) and at 12 weeks postenrollment. We used the protocol of Simon et al [36] to check for depressive symptoms and adverse events. If women scored 5 points above their baseline PHQ-9 score at any time, or if they responded positively to the PHQ-9 item on thoughts of self-harm, then a risk assessment was conducted. This included asking about suicidal thoughts/thoughts of self-harm, frequency of thoughts, triggers, plans, means, lethality, intention, protective factors, and history of suicide attempts/self-harm behavior.

Measures were administered via online questionnaire or by telephone. Demographic data (eg, age, parity, education, income) were gathered both during telephone assessment and in the online battery of baseline questionnaires. The key depression measures (described subsequently) are well validated and their psychometric properties are well described in general populations. The performance and psychometric properties of both generic and perinatal-specific depression measures used in the context of perinatal populations have been recently reviewed and were the subject of a full health technology report in 2009 [42,43].

Before completion of baseline questionnaires and at 12 weeks postenrollment, trained diagnostic interviewers conducted the SCID-IV [38] by telephone to determine a DSM-IV diagnosis of major depression or minor depression [44]. Diagnostic interviewers were blinded to treatment allocation at the 12-week time point.

At baseline and at 12 weeks postenrollment, severity of depressive symptoms was measured online using the revised Beck Depression Inventory (BDI-II) [45]. The BDI-II is a widely used, well-validated, 21-item clinical instrument that measures cognitive, affective, and physiological factors to assess severity of depression. The BDI-II has been used in many studies of perinatal depression [46-50] and has been validated against gold-standard diagnostic criteria in perinatal populations [47].

The primary outcomes were changes in depressive diagnostic status (assessed by the SCID-IV) and depression symptom severity (BDI-II). The categorical outcome (ie, diagnostic status) was analyzed using a contingency table and chi-square test. The continuous outcome (ie, BDI-II) was analyzed in general linear models, which accounted for baseline values as a covariate. Continuous repeated measures data (PHQ-9) were analyzed in mixed-effects growth models using restricted maximum likelihood. Log-likelihood ratio and Akaike information criterion were used to assess model fit.

Consistent with CONSORT standards [58,59], all primary analyses involved planned contrasts of the MumMoodBooster condition versus the TAU control condition with all randomized participants analyzed in their allocated treatment condition. For anxiety and stress symptom severity (DASS-21 Anxiety scale and DASS-21 Stress scale), because transformation failed to improve normality, the nonparametric Mann-Whitney U test was used for between-group comparison. Effect sizes were expressed as Cohen’s d [60] with 95% confidence intervals.

There was some missing data (<5%) on the primary outcomes. Little’s missing completely at random (MCAR) test [61] revealed that these data were missing completely at random (χ² ₃₃₅ ⁼303.56, P=.89). Given the minimal and random nature of missing data, the highly conservative “worst-case” imputation method [41] was used for intention-to-treat analysis. This method replaces missing outcome values in the intervention and control conditions with the least favorable and most favorable scores observed within those respective conditions [41], thus providing a sensitivity analysis of the robustness of the observed treatment effect to type I error [62]. Results are presented on the basis of the intention-to-treat analyses. Computations were executed using IBM SPSS Statistics version 22 (IBM Corp, Armonk, NY, USA).

For the outcome measure of depressive symptom severity (BDI-II), data from a previous trial of CBT [28] for PND provided relevant estimates of variability in baseline scores (mean 23, SD 8.1 points). On this basis, a difference of 6.5 points would be necessary to move mean scores from the “moderately severe” to the “minimal” category of depressive symptoms specified by Beck and colleagues [63] (cut-off between these categories=17 points). We considered this to be the minimum clinically important difference in continuous scores of depression severity [28]. With a power of 0.8 at α=.05, the required n=15.7(8.09/6.5)²=24.3, which rounds to 25 per group. Thus, we aimed for a total sample size of N=50 to achieve sufficient power to detect a clinically important difference in the primary measure of depressive symptom severity.

A total of 178 women began the study registration process (see Figure 3). By the end of the recruitment period, 43 mothers were randomized (24.2% of registrants) to either the MumMoodBooster condition (n=21) or the TAU condition (n=22). Figure 3 details the reasons for attrition between registration and randomization. Twelve weeks following enrollment, two women in the MumMoodBooster condition failed to complete online assessment questionnaires and telephone diagnostic interviews (all women in TAU completed the 12-week assessments).

Women in the two conditions appeared comparable in terms of baseline characteristics (Table 1). As recommended by the CONSORT Statement [58], no significance tests of between-group differences were conducted on baseline data.

In the MumMoodBooster condition, 20 of 21 women (95%) were diagnosed with current major depression and 1 of 21 women (5%) with minor depression. In the TAU condition, 20 of 22 women (91%) met diagnostic criteria for current major depression and 2 of 22 women (9%) for minor depression (Table 1).

The mean age of mothers was close to the most recently available Australian national average (30 years [64]) and their infants averaged just over 6 months of age. The sample contained a higher proportion of women born in Australia than the latest national average (national average=70%) and a lower proportion of first-time mothers (national average=43%). The mean EPDS at screening was 16.2 (SD 2.9), similar to the mean observed in other Australian samples of depressed perinatal women [47]. As reported in studies of major risk factors for PND [65], 67.4% (29/43) of the cohort had experienced a past major depressive episode.

In the MumMoodBooster condition, 79% (15/19) of women no longer met DSM-IV criteria for depression at 12 weeks; this was 18% (4/22) in the TAU condition. In the intention-to-treat analysis, a chi-square test revealed a statistically significant between-group difference in the frequency of depressive diagnosis (major or minor depression) at 12-week follow-up (Yates corrected χ² ₁=10.3, P=.001). Figure 4 shows the proportions on which the intention-to-treat analysis was conducted.

Similarly, women in the MumMoodBooster condition also showed reduced severity of depression symptoms, whereas those in the TAU condition displayed little improvement (see Table 2). After controlling for baseline scores, mean depression symptoms on the BDI-II in the intervention group at 12 weeks were significantly lower than that of the TAU group (P=.01) and this represented a large effect size (Table 2). Results for both diagnostic status and BDI-II scores were also statistically significant in observed-case analyses before imputation.