Study sample
Participants (N = 612) were recruited from seven WIC offices in Michigan. WIC, a federally funded program, is the leading and largest public health nutrition program in the United States. In 2014, WIC nationwide served nearly 9.3 million low-income pregnant, postpartum, and breastfeeding women and children (0–5 years) [22]. WIC provides services to individuals with annual household income at or below 185% of the federal poverty line. Recruitment took place from September 2012 to January 2015. Recruiters (who were peers of the study participants) personally invited women who came to our collaborating WIC offices for their WIC appointment to be screened. Potential participants filled out a screening pencil-and-paper survey; then the peer recruiters measured their height and weight to calculate body mass index (BMI). To be eligible to participate, women were required to be not pregnant, non-Hispanic Black or white, overweight or obese (BMI 25.0–39.9 kg/m2), between 6 weeks and 4.5 years postpartum, 18–39 years old, free of type 1 or 2 diabetes, and able to walk more than 1 block without resting. Peer recruiters obtained written consent forms from all eligible participants. After completing a baseline interview via phone (T1), consented women returned to the WIC office where they had been recruited to be randomized to either an intervention group (410 participants) or comparison group (202 participants). The intervention participants then received ten intervention video lessons in DVD format in person and were told not to share any DVDs with anyone at any time. The comparison group received printed materials on stress management, healthy eating, and physical activity from credible websites, including some government websites (e.g., U.S. Department of Agriculture). WIC staff at our collaborating sites were not aware of participants’ randomization assignments. Detailed descriptions of recruitment and study procedures have been described elsewhere [23, 24]. Michigan State University and the Michigan Department of Community Health Institutional Review Board approved the study procedure.
Intervention
Theoretical framework
Social Cognitive Theory (SCT) was used to guide the intervention design. SCT emphasizes that observing peer role models making positive behavioral changes motivates and empowers individuals to engage in positive behaviors [25]. The central concept of this theory is reciprocal determinism, which means that there is a reciprocal interaction of personal factors (e.g., self-efficacy and emotional coping response), environmental factors (e.g., social support), and behaviors (e.g., lifestyle behaviors) [25]. Our intervention addressed personal and environmental factors to promote stress management and healthy lifestyle behaviors, with the ultimate goal of preventing further weight gain. Personal factors addressed were self-efficacy (one’s confidence to perform a specific behavior) and emotional coping response. The environmental factor addressed was social support.
Description of intervention
A detailed description of the intervention video development has been previously published [23]. Each intervention video featured four peers of the target audience (hereafter referred to as featured women) and their family members, especially young children. The featured women modeled positive behavioral changes by demonstrating practical skills to overcoming common daily social, psychological, and environmental challenges to manage stress better, eat healthier, and be more physically active. These women and their young children were filmed at home, in a local grocery store, and in their neighborhoods over a period of one year so that we could show how they made positive changes over time. To help participants build confidence (self-efficacy), our intervention helped them identify strengths, for example, recognizing their existing skills and making small steps to change. To improve emotional coping response, we provided, for example, effective ways to identify triggers of negative emotions and to respond to those triggers. To increase social support, our intervention explained helpful strategies, for example, selecting positive social support persons and eliciting and building social support. There were four stress management video lessons. The first video lesson focused on better ways to handle daily hassles (e.g., use 5Ws [who, what, when, where, and why] and H [how]) to identify root causes of a problem, be a good mom rather than a supermom, be consistent and stay calm when feeling stressed out). Lesson two covered time saving tips for busy moms (e.g., say no and set priorities, get some tasks done at night instead of early morning, and manage time with a to-do list). Lesson three covered effective ways to handle negative feelings (e.g., breathe deeply, count to ten to stay calm, remove oneself from a troubling situation for a moment, and speak positively to oneself). Lesson four included effective ways to help with parenting (e.g., listen and talk to children, be consistent, get down to the child’s eye level to communicate with the child, and use a responsibility chart). A detailed description of lessons five to ten (healthy eating and physical activity) has been published elsewhere [21, 23].
Intervention implementation
A detailed description of intervention implementation has been previously published [23]. During the 16-week intervention, participants watched ten video lessons (20 min per video lesson) in DVD format at home weekly (weeks 1–4: stress management) followed by every other week (weeks 5–16: healthy eating and physical activity). After watching a designated video lesson, participants circled responses on a worksheet that asked about content of the designated video lesson watched, then mailed the worksheet to the study office using a self-addressed stamped envelope. We used the returned worksheet as an indication of watching the video lesson. Participants also joined peer support group teleconferences (30 min per session, weekly for the first four weeks, then every other week for weeks 5–16) led by moderators who were peer educators or WIC dietitians trained in motivational interviewing and group facilitation skills. We recorded attendance based on whether a participant joined a particular peer support group teleconference.
Measures
Self-report data were collected through telephone interviews at three time points: baseline (T1), the end of the 16-week intervention (T2), and three-month follow-up (T3). We used assessments developed for the target population that have been shown to demonstrate construct validity and good internal reliability (α) to measure self-efficacy, emotional coping response, and social support [26].
Self-efficacy
The survey used to measure self-efficacy had ten items (α = 0.92) and asked about participants’ confidence in managing stress [26]. For example, “You can relax, even when your kids scream.” Response options were on a four-point scale ranging from 1 = not at all confident to 4 = very confident. The overall self-efficacy score was the mean of the ten-item scores, with a higher score indicating higher self-efficacy.
Emotional coping response
The survey used to measure emotional coping response had five items (α = 0.91) [26] and asked participants about strategies used to cope with stress. For example, “How often do you deal with or prevent stress by taking a walk?” Response options were on a four-point scale ranging from 1 = rarely or never to 4 = usually or always. The overall emotional coping score was the mean of the five-item scores, with a higher score indicating better emotional coping response.
Social support
The survey used to measure social support had six items (α = 0.87) [26]. For example, participants were asked “You can rely on family members, friends, coworkers, or other people for support when you need to ask for advice.” Response options were on a four-point scale ranging from 1 = rarely or never to 4 = usually or always. The overall social support score was the mean of the six-item scores, with a higher score indicating more social support.
Stress
The Perceived Stress Scale (nine items) with established validity and reliability was used to measure stress perception [27]. Participants were asked about their perception of stress in the past month. Response options were on a four-point scale ranging from 1 = rarely or never to 4 = usually or always. The overall stress score was the mean of the nine-item scores, with a higher score indicating lower stress.
Depressive symptoms
The Center for Epidemiologic Studies Depression Scale with established validity and reliability was used to measure depressive symptoms (20 items) [28]. Response options were on a four-point scale ranging from 0 = rarely or none of the time to 3 = most or all of the time. Responses to the 20 items were summed to create a depressive symptom score ranging from 0 to 60, with a higher score indicating more depressive symptoms. A score of 16 was used as a cutoff value for indication of being at risk for clinical depression [28].
Positive and negative affect
The Positive and Negative Affect Scale with established validity and reliability was used to measure affect (18 items) [29]. Seven items measured positive affect and 11 items measured negative affect. Response options were on a five-point scale ranging from 1 = very slightly or not at all to 4 = extremely. The overall positive affect score was the mean of the seven-item scores, with a higher score indicating more positive affect. The overall negative affect score was the mean of the 11-item scores, with a higher score indicating less negative affect.
Statistical analyses
Statistical analyses were conducted using NCSS Version 11 (Kaysville, UT) on 569 women (387 intervention and 182 comparison participants) after excluding 43 women who became pregnant during the study. We performed descriptive analysis on all variables, t-test for continuous variables, and chi-squared test for categorical variables. The outcomes of interest were self-efficacy, emotional coping response, social support, stress, depressive symptoms, and positive and negative affect at T2 and T3. To assess intervention effect, we performed a general linear mixed model for repeated measures, using baseline measures as adjusting covariates. We chose a general linear mixed model over intent-to-treat analysis because our overall retention rate for the phone interview was low for T2 (59%) and T3 (55%). A general linear mixed model is a partial intent-to-treat analysis, a method using all available data without any ad hoc imputation as in the intent-to-treat approach. Simulation studies have demonstrated that analysis with mixed models without any ad hoc imputation provides more powerful tests than mixed model analysis with last observation carried forward or imputation for intent-to-treat analysis [30]. Effect size and 95% confidence intervals were calculated using Cohen’s d.