Social Media-based Parenting Program for Women With Postpartum Depressive Symptoms: An RCT

We conducted the study at 6 urban pediatric practices affiliated with a large children’s hospital from December 2019 to August 2021. These practices provide comprehensive care to a racially and socioeconomically diverse, urban population of infants, children, and adolescents. The practices all screen for PPD using the Edinburgh Postnatal Depression Scale (EPDS) at selected infant well-child visits through 6 months of age, consistent with the American Academy of Pediatrics’ preventive care recommendations.²¹  The EPDS is a validated 10-item scale for PPD with scores >9 having a sensitivity of 97% and specificity of 43% for major or minor depression.²²  The EPDS includes a single item querying respondents regarding suicidal ideation. For those women who screen positive for PPD or suicidality, the practices can provide a community mental health referral. This study was approved by the institutional review board at the Children’s Hospital of Philadelphia and was registered with Clinicaltrials.gov before enrollment of participants (NCT 04045132).

Participants were eligible if they were mothers aged >18 years of age, screened positive for minor or moderate depressive symptoms (EPDS 10–19) at their child’s well visit, could read and write in English, and had access to a smartphone or computer tablet. To assess effects of the intervention on child developmental status, participants were excluded if they had infants that were premature (<35 weeks’ estimated gestational age), identified with genetic syndromes or congenital anomalies that impact child development, or enrolled in early intervention services. Eligible participants completed written informed consent, were enrolled in the study, and were reimbursed for their cell phone data plan.

The main intervention was Parenting with Depression (PWD), a social media-based parenting program developed for new mothers with PPD symptoms to enhance their parenting skills and improve parent–child interactions. PWD was administered through a Facebook secret group platform to ensure confidentiality of participants outside the group. Participants were encouraged to “friend” and respond to other member posts. PWD consisted of 8 weekly topics: depression psychoeducation and behavioral activation, infant temperament (personality), play, feeding, safety, sleep, parent–child interactions (laughter), and shared book reading.²³  Content for each topic was released separately in 3 postings onto the Facebook site over the course of a week and consisted of video vignettes, narrated PowerPoint presentations, and written materials for ease of posting. A facilitator reviewed posts daily, commented on participants’ postings, provided coaching tips, and removed any inappropriate postings.

Participants were stratified by practice site and randomized 1:1 by the study biostatistician in random blocks of size 2 to 4 to mask guessing of the next assignment to the PWD program plus MoodGym, an online cognitive behavioral therapy program, or MoodGym alone. MoodGym is an evidence-based, online cognitive behavioral treatment program that has been shown effective at reducing mild-to-moderate depression symptoms.²⁴  In addition, all participants received a list of community mental health resources. Mothers in the PWD plus MoodGym group were enrolled in the Facebook group in sizes of 6 to 10 participants to facilitate group communication.

At baseline, participants completed measures of demographics (age, sex, race/ethnicity, maternal education level, family structure, family income) and social support. To monitor depression severity and assess for suicidality, participants completed the EPDS and an item from the National Comorbidity Survey concerning any community mental health service use each month for 3 months.²⁵  After enrollment, women with EPDS scores >20 or who reported suicidality (EPDS question #10) were contacted by a psychologist to assess risk, were placed on a weekly call list to assess for worsening risk, and provided with assistance in obtaining urgent mental health referrals if they were not already in treatment. Those deemed to be at high risk were referred for emergent mental health treatment and were excluded from further participation until their EPDS scores were <20 and they were not suicidal. Social support was measured by the Multidimensional Scale of Perceived Social Support, a 12-item scale that assesses perceived social support from family, friends, and a significant other.²⁶ 

Parenting measures were completed at baseline and at follow-up 3 months later using a Research Electronic Data Capture database with blinded assessment by study staff. Measures consisted of the Parenting Sense of Competence Scale (PSOC), the Parenting Stress Index-Short Form (PSI-SF), and the Parent–Child Early Relational Assessment (PCERA). The PSOC is a validated 17-item self-report measure of parenting self-esteem and competence and consists of 2 factors: satisfaction and efficacy.²⁷  The PSI-SF is a validated 36-item measure of parenting stress.^28,29  The PCERA is a validated 65-item (29 parental, 28 child, and 8 dyadic items) videotape assessment designed to measure the quality of affect and behavior in parent–child interactions.^30,31  The PCERA has demonstrated good interrater reliability, internal consistency, discriminant and concurrent validity, and sensitivity to change.^31–34  We focused on 3 PCERA subscales: (1) parental positive affective involvement and verbalization, (2) parental negative affect and behavior, and (3) parental intrusiveness, insensitivity, and inconsistency. An external rater trained and experienced in PCERA coding scored all videotapes. To ensure reliability in coding, a second rater coded 28 (20%) tapes and overall agreement was 89%. We determined the level of engagement of PWD participants with the program by the number of separate posts in the Facebook group using Google Analytics. We categorized engagement as low 0 to 8 posts, moderate 9 to 16 posts, or high 17 to 24 posts.

We examined summary statistics and assessed differences between groups using the t test for continuous variables and χ² test for categorical variables. We considered a P value <.05 for statistical significance. Our primary outcome was change in the 3 PCERA parenting subscales among groups from baseline to follow-up. Secondary outcomes included changes in EPDS, PSOC, and PSI-SF scores. Because there were small differences in follow-up time among participants and to accommodate dropout, we developed linear mixed effects models to assess difference-in-differences in outcomes between groups while accounting for clustering within practices and adjusting for imbalances in baseline characteristics. Primary analyses were conducted using intention-to-treat analysis in which groups were analyzed by their original randomization assignments and missingness was assumed to be completely at random. We used group X time interactions to assess differences in outcomes. On the basis of a 2-sample test of means, a sample size of 75 participants yielded an 80% power to detect an effect size between 0.79 and 0.96 among groups, assuming an interclass correlation coefficient between 0.01 and 0.1 accounting for correlation among participants of the same practice sites and 20% attrition. We conducted exploratory analyses of PCERA scores among participants in the PWD group by level of engagement with the PWD program, by whether additional study resources for severe depressive symptoms or suicidality were provided, and by report of any community mental health treatment during the study period.

Of 175 potentially eligible mothers referred to the study (Fig 1), we excluded 100 for the following reasons: didn’t meet eligibility criteria (n = 26), couldn’t be reached or declined participation (n = 41), withdrew before randomization (n = 1), or expressed interest in participating after we reached our enrollment goal (n = 32). Thus, 75 participants were enrolled in the study: 38 in the intervention group (PWD + MoodGym) and 37 in the comparison group (MoodGym alone). Sixty-six (88%) of the 75 participants completed the 3-month study.

Participants were predominantly Black, non-Hispanic race (69%), single or separated (57%), with greater than a high school education (66%), and with a family income <$55 000 (68%). Mothers in the intervention group were more likely to report having some college or college degree (76% vs 54%, P = .04) than those in the comparison group (Table 1). There were no other demographic differences among groups.

At baseline, there were no differences in PCERA factors 1 to 3 scores among groups (Table 2). In addition, there were no differences in PSOC, PSI-SF, Multidimensional Scale of Perceived Social Support, and EPDS scores among groups. The average EPDS scores indicated mild-to-moderate depressive symptoms at baseline.

Mean unadjusted EPDS scores decreased more rapidly in the intervention group compared with the comparison group (Table 3 and Fig 2a). However, by the conclusion of the 3-month study period, mean unadjusted EPDS scores were in the minimal-to-mild depression severity range and were similar in both groups. A substantial number of participants (49% in the comparison group versus 37% in the intervention group) received additional study resources because of severe depressive symptoms or suicidality. In addition, a large proportion of participants (43% comparison group versus 40% intervention group) received community mental health services during the study period (Table 3).

For the main outcomes of PCERA factors 1 to 3, there were no differences in adjusted baseline scores (Table 2). Over the 3-month study period, PCERA scores increased over time in both groups, suggesting minor improvements in parent–child interactions. There were no significant group X time interactions; PCERA scores improved similarly in both groups (Table 4). There was a site difference in which site 5 demonstrated greater baseline PCERA scores than other sites; thus, we controlled for site in the analysis. There were no adjusted differences in PCERA scores by caregiver education.

For secondary outcomes, adjusted EPDS scores demonstrated a more rapid decline in the intervention group relative to the comparison group (Fig 2a). Differences in adjusted EPDS scores were significant at 1 month (−2.9, 95% confidence interval [CI], −4.8 to −1.0) and 2 months (−2.0, 95% CI, −3.9 to −0.15), but not at 3 months (−1.2, 95% CI, −3.2 to 0.7). Both groups demonstrated improvement in adjusted PSOC scores and declines in adjusted PSI-SF scores over time, but there were no differences between groups in changes in these measures over time (Fig 2b–c).

In exploratory analyses, we examined PCERA factor 1 to 3 scores among intervention participants by the categorized level of their engagement with the PWD program. Those participants who demonstrated moderate (N = 5) or high (N = 4) levels of engagement were few but had greater improvements in PCERA factor 1 (0.24 or 0.39 vs −0.01) and factor 3 (0.35 or 0.5 vs 0.18) but not factor 2 (0.04 or 0.20 vs 0.13) scores than those with low levels of engagement (N = 23). Similarly, those participants who reported any community mental health treatment (N = 13) demonstrated greater improvement in PCERA factor 1 (0.13 vs 0.04), factor 2 (0.32 vs −0.02), and factor 3 (0.36 vs 0.15) scores than those who did not report any treatment (N = 19). Finally, those who received additional study resources had lesser improvements in PCERA factor 1 (0.16 vs 0.21) and factor 3 (0.26 vs 0.40) but not factor 2 (0.27 vs 0.14) scores than those who did not report severe depressive symptoms or suicidality. Given the small sample sizes, we did not perform tests of statistical significance for these exploratory analyses.

Three participants (2 intervention and 1 comparison group) were hospitalized for worsening depression/suicidality during the study period. These adverse effects were determined to be unrelated to the interventions.

In this study, the use of a social media-based parenting program for women with PPD symptoms combined with online cognitive behavioral treatment resulted in a more rapid decline in depressive symptoms than online cognitive behavioral treatment alone, but results were similar at 3 months. This result is consistent with a systematic review and meta-analysis of 48 randomized controlled trials of group-based parenting programs that found short-term improvements in maternal depressive symptoms for up to 6 months.³⁵  The trials reported in the systematic review consisted of in-person group programs. Our study is among the first to examine the effects of a parenting program administered through a social media platform. This has important implications for program implementation during a pandemic and can address barriers to participation in this population.

The mechanism for how PWD and other parenting programs improve maternal depressive symptoms isn’t clear. One hypothesis is that parenting programs improve parent–child dyadic functioning and this in turn improves parental psychopathology.³⁵  In our study, we found similar improvements in parental responsiveness in both groups, so this hypothesis wouldn’t explain the differential, more rapid improvement in depressive symptoms in the PWD group. Another hypothesis is that PWD and other similar programs have a direct effect on maternal functioning through improvements in parenting self-efficacy and parenting stress. We were unable to test this hypothesis, because we only measured parenting self-efficacy and parenting stress pre- and postintervention and not monthly as with the EPDS.

The results of the study are consistent with our previously published pilot study results.²³  In this previous study, we found that women who participated in the online PWD program reported moderate-to-large improvements in depression and parenting competence scores. However, in the current study with a larger sample size, we found similar improvements in depression scores but only modest improvements in parenting competence scores. In addition, the comparison group reported similar improvements in these measures in the current study but not in the previous study.

For our main outcome, we found no difference in changes in our measure of responsive parenting among groups. Both groups demonstrated modest improvements in responsive parenting. This may have been because of improvements in maternal mood in both groups. A large proportion of participants in both groups (41%) reported accessing community mental health services during the study period, which may have led to the observed improvements in mood, particularly for the comparison group. Any separate effect of the PWD program on parenting outcomes among groups may have been overwhelmed by the beneficial impact of mental health services. In support of this hypothesis, we found in exploratory analysis that those PWD group participants who reported any mental health services during the study period demonstrated greater improvements in responsive parenting than those who did not.

It is important to point out that this study occurred during the height of the COVID-19 pandemic, when maternal mental health distress was high nationally.²⁰  We purposefully selected women with mild-to-moderate depressive symptoms and excluded those with severe depressive symptoms or high suicide risk for study eligibility. However, 41% of study participants (46% in control group versus 38% in intervention group) later experienced severe depressive symptoms (EPDS >20) or suicidal behavior during the study period that necessitated safety evaluations, immediate mental health referrals if indicated, and provision of community mental health resources. All but 2 of these women (94%) later reported any mental health treatment. This high level of worsening depression and mental health services use is in contrast to the low level of mental health services use prepandemic among women with PPD symptoms when virtual mental health services were not routinely provided.³⁶ 

Another potential effect of the COVID-19 pandemic may have been less engagement with the social media-based parenting program. Only 9 (28%) of the intervention group participants viewed and posted comments to at least one-third of the weekly content. In our pilot study, 73% of participants did so.²³  In an exploratory analysis, we found that women who engaged this much had greater improvements in responsive parenting than those with less engagement. This suggests that passively viewing content is not as helpful as engaging in the exercises and interacting with other participants and the study facilitator.

There are limitations to our study. First, the study occurred in a single urban geographic area with a predominantly Black population. Findings may not be generalizable to other geographic regions and to other racial and ethnic minority populations. Second, the study occurred during the height of the COVID-19 pandemic. Any beneficial impact of the PWD program may have been overwhelmed by the high levels of maternal distress during the pandemic and the high level of community mental health services use as discussed above. Third, engagement with the PWD program was lower than expected. Those who demonstrated greater engagement with the program also exhibited greater gains in responsive parenting.

Participation in a social media-based parenting program led to a more rapid decline in PPD symptoms but no differences in responsive parenting, parenting stress, or parenting competence relative to a comparison group with access to an online cognitive behavioral therapy program. Social media may be an important platform to provide parenting support for women with PPD symptoms when in-person participation is difficult, but greater attention to participant engagement and ensuring and accounting for mental health treatment are needed to improve parenting outcomes with social media-based parenting programs.

We thank the Pediatric Research Consortium at the Children’s Hospital of Philadelphia and the staff and clinicians at participating practices for their support of this study.

CI

confidence interval

COVID-19

coronavirus disease 2019

EPDS

Edinburgh Postnatal Depression Scale

PCERA

Parent–Child Early Relational Assessment

PPD

postpartum depressive

PSI-SF

Parenting Stress Index-Short Form

PSOC

Parenting Sense of Competence

PWD

Parent With Depression