We conducted a 2-arm, parallel-group randomized controlled trial among members of Kaiser Permanente Northern California (KPNC), an integrated health care delivery system that provides care for over 66,000 pregnant and postpartum individuals annually. PPD screening is part of standard postpartum care at KPNC [32,33] and also occurs at each pediatric well-baby visit. Results of all PPD screening and diagnosis are captured in KPNC’s comprehensive electronic health record database.

This study was approved by the KPNC Institutional Review Board (approval number 1548855) and registered with ClinicalTrials.gov on June 1, 2020 (ID: NCT04414696). All participants provided documented informed consent and could withdraw from the study at any time. All data are deidentified. Participants received gift cards as incentives after completing each assessment: US $20 after the baseline and 3-month assessments, and US $30 after the 6-month assessment (including the infant ASQ screener), for a total of US $70 in incentives.

From November 2020 through September 2022, 12,269 postpartum individuals who met the following inclusion criteria in the electronic health record were invited to participate in the study via letter and email invitations: 2‐6 months postpartum, no current depression diagnosis, and at high risk for PPD (identified using PPD screening PHQ-9 score 10‐19 or PHQ-2 score ≥3, or history of depression diagnosis and antidepressant medication prescription prior to delivery date). Additional eligibility criteria (including reassessment of depression symptoms using a PHQ-8 screener) were assessed using online questionnaires prior to the baseline survey. See Table S1 in Multimedia Appendix 1 for detailed inclusion and exclusion criteria.

Eligible participants completed informed consent documents and baseline surveys through the REDCap website (Research Electronic Data Capture; Vanderbilt University). After survey completion, participants were mailed an accelerometer and were asked to wear the accelerometer for 7 consecutive days, 24 hours per day, and then mail it back. Participants who returned the accelerometer were eligible to be randomized.

Participants were randomized by the study project manager into the intervention group or control (usual care) group using minimization, implemented using QMinim [34]. Minimization aims to achieve between-group balance in distributions of pre-specified important factors associated with outcomes of interest. Variables included in the randomization procedure included number of children at home (1, 2+), self-reported race and ethnicity category (Asian/Pacific Islander, Hispanic, non-Hispanic Black, non-Hispanic White, other), baseline PHQ-8 score (Patient Health Questionnaire-8; 10-14, 15-19), and PA level prior to pregnancy (below recommendations, at or above recommendations). Investigators and data analysts were blinded to the assigned study group. Research assistants, who informed participants of their assigned study group, and participants were not blinded.

In addition to usual postpartum care, participants randomized to the intervention group received access to MomZing (Klein Buendel), an eHealth PA intervention tailored to postpartum individuals [35]. MomZing is an online library of 98 ten-minute exercise videos that was developed with and for postpartum individuals. The exercise videos on MomZing instructed postpartum individuals how to safely exercise with their baby based on the infant’s weight and developmental stage, did not require exercise equipment, and featured postpartum individuals (who were not fitness instructors) exercising with their own infant. The website also included a page to set a weekly PA goal and visually tracked (via a graph) the number of videos watched per week. Participants randomized to the intervention group received individual login information for the MomZing website via email and up to 6 reminders (text and email per each participant’s preference) to log into the MomZing website (73% text and 27% email).

Participants randomized to the control group received usual postpartum care for women at increased risk of depression, which typically is a brief discussion about their depression symptoms with their obstetric provider.

Follow-up surveys were sent via email at 3 and 6 months postrandomization. After completion of each survey, participants were mailed an accelerometer and were asked to wear the accelerometer for 7 consecutive days, 24 hours per day, and then mail it back. An additional assessment occurred when the participant’s child was 12 months old; for some participants, this occurred simultaneously with the 6-month follow-up surveys. Follow-up data collection ended in April 2023.

Primary outcomes were depressive symptoms and device-measured moderate-to-vigorous intensity PA (dm-MVPA), measured at 3 months postrandomization. Depressive symptoms were assessed using the validated 8-item PHQ-8 screener [36,37], with higher scores indicating greater symptom severity (range 0‐24). dm-MVPA was measured using the ActiGraph GT3X+ accelerometer (ActiGraph, Pensacola), worn on the non-dominant wrist. Raw accelerometer data were downloaded using ActiLife software (ActiGraph, Pensacola) and counts were summarized for each 1-second and 60-second epoch. Wear time was identified using the Choi algorithm [38-40], and wake (non-bedrest) was identified using the Tracy algorithm [41]. Individuals with at least 4 days (including at least 1 weekend day) of at least 10 hours of waking accelerometer wear were considered to have sufficient accelerometer data for analyses [42]. Waking wear time was categorized into intensity categories using the Hibbing 2-regression model, which was developed for wrist-worn accelerometer data [43]. Average dm-MVPA duration was calculated across valid days.

Secondary outcomes included self-reported moderate-to-vigorous intensity PA (sr-MVPA) at 3 and 6 months postrandomization, and depressive symptoms and dm-MVPA at 6 months postrandomization. sr-MVPA duration was assessed using the sports and exercise domain of the Pregnancy Physical Activity Questionnaire [44]. In addition, we created a subscale of the sports and exercise domain of the Pregnancy Physical Activity Questionnaire (sr-MVPA-MomZing) limited to activities included in MomZing videos: dancing; yoga or Pilates or stretching or core strengthening exercises, exercise videos, or computer games; aerobic exercise videos or computer games; and weightlifting and resistance exercises (ie, MomZing included workouts where the infant was lifted or carried).

Additional outcomes were measured using validated questionnaires and included sleep quality, anxiety symptoms, and perceived stress at 3- and 6- months postrandomization, mother-infant bonding at 3 months postrandomization, and infant development at 12 months of age. Sleep quality was measured using the Pittsburgh Sleep Quality Index [45], with higher scores indicating worse sleep quality (range 0‐21). Anxiety symptoms were measured using the GAD-7 screener [46,47], with higher scores indicating greater symptom severity (range 0‐21). Perceived stress was measured using the Perceived Stress Scale [48,49], with higher scores indicating higher stress (range 0‐40). Mother-infant bonding was assessed using the Mother Infant Bonding Scale [50], with lower scores indicating better bonding (range 0‐24). Infant development across 5 domains (communication, gross motor function, fine motor function, problem-solving, and personal social skills) was assessed using the Ages and Stages Questionnaire-3 for 12-month-old infants [51], with higher scores indicating more typical (“on schedule”) development (range 0‐60).

Participants reported injuries or illnesses related to exercise at 3 and 6 months postrandomization.

Participants allocated to the intervention group received a feedback survey at 3 months postrandomization that assessed their use of the MomZing website, reasons for not using the website, experience and satisfaction with the MomZing intervention, and qualitative feedback about the intervention. The satisfaction questions were adapted from a validated satisfaction questionnaire for web-based health interventions with Likert scale responses, with higher scores indicating higher satisfaction (range 12‐60) [52].

Original sample size calculations were based on a planned sample size of 100 participants per group. With our achieved sample size of 99 participants (50 in the intervention group, 49 in the control group), 80% power, and 5% type I error rate, the minimum detectable difference in means for depressive symptoms and dm-MVPA was 0.57 standard deviation units, which is considered a “medium” effect size [53]. Using the observed outcome distributions (Table 1), this corresponds to 3.2 units for PHQ-8 score and 24.1 minutes for dm-MVPA.

We compared baseline characteristics between the intervention and control groups and between the entire intervention group and the subset of the intervention group who engaged with the intervention (watched at least 1 MomZing video).

Primary analyses followed an intent-to-treat approach. We used multivariable linear regression models, adjusted for variables used in the randomization procedure (number of children at home, race and ethnicity, baseline PHQ-8 score, pre-pregnancy PA level) and corresponding baseline value of the outcome, to estimate mean differences in each outcome associated with the MomZing intervention.

Secondary analyses followed a modified intent-to-treat approach. We excluded participants allocated to the intervention group who did not engage with the intervention (did not watch at least 1 video on the MomZing website during the 3-month period postrandomization).

We used multiple imputation using chained equations to impute missing outcome and baseline MVPA data with 10 imputations. Variables included in imputation models are shown in Table S2 in Multimedia Appendix 1. Regression models were run using each imputed dataset. Results were combined using Rubin’s rules [54]. We also conducted complete case analysis as a sensitivity analysis. Analyses were conducted in SAS (version 9.4; SAS Institute Inc.) and R (version 4.3.1; R Foundation for Statistical Computing).

Of the 99 randomized participants, 27 (54%) and 26 (53%) in the intervention and control groups, respectively, completed all 3-month follow-up activities (online surveys and accelerometer; Figure 1). At baseline, participants were 4 months postpartum, on average, and had moderately severe depressive symptoms (mean PHQ-8=12.6; Table 2). Most participants were members of a minoritized racial or ethnic group (63%), not currently working (74%), married or living with their partner (86%), did not meet PA recommendations prior to pregnancy (61%), had a college education or more (80%), and had 2 or more children at home (59%). Baseline characteristics were similar between the intervention and control groups (Table 2).

In total, 52% (n=26) of participants allocated to the intervention group logged in to the MomZing website and watched at least 1 video. The median minutes of videos watched per week peaked at 10 minutes 2 weeks after randomization, then fell to zero for the rest of the follow-up period (Figure 2). Baseline characteristics were similar between participants who engaged with the intervention (watched at least 1 MomZing video) and participants in the control group . Within those allocated to the intervention, characteristics differed somewhat by engagement with the intervention: compared to participants who did not engage with the intervention, those who did were fewer months postpartum, more likely to be non-Hispanic White, more likely to have a household income of US $100,000 or greater per year, less likely to be working, and less likely to meet PA recommendations in pre-pregnancy (Table S3 in Multimedia Appendix 1).

Findings from modified intent-to-treat analyses were similar to those from intent-to-treat analyses for primary, secondary, and additional outcomes (Tables 1, 3 and 4).

Findings from complete case analyses were similar to those using imputation (data not shown).

There were no adverse events reported during the study.

Among those who reported using the MomZing website (44%, n=22), the mean satisfaction score was 40, indicating moderately high overall satisfaction with the intervention (Table 5). In total, 90% rated MomZing as good or very good or excellent. Over 60% of participants who reported using the MomZing website found it very or moderately helpful for being more physically active. Qualitative feedback also indicated that participants wanted more frequent reminders to use the MomZing website, more intense workouts, and an app that is accessible on multiple devices. The most common reason for not using the MomZing website was lack of time (77%).