Participants and procedure
The current analyses are part of a larger study examining psychosocial factors that relate to body image and disordered eating during the postpartum period. However, to not bias recruitment sampling, women were recruited to take part in a survey of postpartum women’s experiences broadly (with no emphasis on body image or disordered eating) from November 1, 2019, through June 6, 2020. Women, age 18–39, who either gave birth since 2019 or received an annual Well-Woman gynecological check-up within the University of North Carolina at Chapel Hill Healthcare system were e-mailed a recruitment letter with the survey link. For those with no email address listed, letters were mailed to addresses on record. The mailing lists included 937 women potentially eligible for the postpartum group and 3000 women potentially eligible for the control group. Additionally, recruitment flyers were posted in local health settings, ads were shared on social media, and emails were sent over listservs. Women were eligible if they self-reported as having been pregnant and given birth within the past 12 months (postpartum group) or if they had never been pregnant (control women), were age 18–39, and identified as female. Postpartum women did not have to be primipara and could have older children of any age. Additionally, they were not excluded if they experienced miscarriage, infant loss, or terminated their pregnancy. Given the focus of the study on postpartum women who have already given birth, pregnant women were excluded from participating. However, they were invited to recontact the research team after they gave birth if they were interested in participating.
Given the study link was shared on all recruitment materials (i.e., emails, letters, flyers), we cannot know how many potential participants did not self-select to participate, including those with incorrect addresses or emails. A total of 1565 respondents clicked on the survey link; however, 812 individuals did not make it past the initial screening questions and study instructions (potentially due to lack of interest and/or ineligibility). Of the 753 total eligible participants (both postpartum and control women), 306 postpartum women and 153 control women completed the study, resulting in a completion rate of approximately 61% (n = 459).
Participants provided electronic consent before completing the survey which included questions about mental health symptoms. As compensation, they had the option to enter into a gift card drawing. The study was approved by the institutional review board of UNC.
In total, 306 postpartum and 153 control women completed the study. Given the focus on mental health during the pandemic, we focused on 232 (75.8%) postpartum and 137 control women (89.5%) who participated after the WHO declared a global pandemic to address aim 1, comparing postpartum and control women on symptoms. Additionally, given that 2019-nCOV distress items were added to the survey on April 3, 2020, of those that participated during the pandemic, only 89 (38.4%) postpartum and 110 (80.3%) control women received these items and were included in addressing aim 2, examining the relationship between 2019-nCOV distress and mental health symptoms. Demographic information for participants who participated during the pandemic are presented in Table 1. In terms of race and ethnicity, about 92.7% of postpartum women identified as White (n = 215), 3.4% as biracial or multiracial (n = 8), 1.7% as Black (n = 4), and 0.4% as Asian (n = 1). Among control women, 83.2% identified as White (n = 114), 5.8% as Asian (n = 8), 5.1% as Black (n = 7), 3.6% as biracial or multiracial (n = 5), and 1.5% as other (n = 2). These demographics are slightly less diverse than the local population according to the 2010 US Census with approximately 72% of residents identifying as White (United States Census Bureau 2010). Postpartum women gave birth an average of 21.97 weeks (SD = 14.01; range 1–52) prior to participation, and approximately 62.4% were primipara.
Self-reported demographic data for age, level of education, race, and ethnicity were collected via a set of questionnaires created for this study. Body mass index was calculated using self-reported height and weight. Postpartum women provided time since birth (weeks).
Mental health history
Participants indicated if they had ever sought treatment for or been diagnosed with depression (including postpartum depression), an anxiety disorder (including OCD), or an eating disorder.
The Center for Epidemiological Studies Depression Scale (CES-D; Radloff 1977) measured frequency of depressive symptoms in the past week with a clinical cutoff of 16 indicating a likely current depressive episode (Weissman et al. 1977). The CES-D has good internal consistency (α = 0.88) and discriminant validity (rs = − 0.43–0.57; Knight et al. 1997) and has been normed in perinatal samples (Campbell and Cohn 1991; Logsdon and McBride 1994; Marcus et al. 2003). Coefficient alpha for the current sample was 0.91.
The Anxiety subscale of the Depression Anxiety Stress Scale-21 (DASS-21; Lovibond and Lovibond 1995) assessed symptoms associated with physical arousal in response to fear in the past week. The anxiety subscale creates a score comparable to the full DASS-42 (Lovibond and Lovibond 1995). The DASS-21 has demonstrated good convergent validity (ηp2 = 0.21) and reliability (α = 0.78; Norton 2007). Coefficient alpha for the current sample was 0.79.
The Dimensional Obsessive–Compulsive Scale (DOCS; Abramowitz et al. 2010) assessed severity of obsessive–compulsive disorder-type (OCD) symptoms in the last month across four dimensions: Concerns about Germs and Contamination; Concerns about being Responsible for Harm, Injury or Bad Luck; Unacceptable Thoughts; and Concerns about Symmetry, Completeness, and the Need for Things to be “Just Right.” The DOCS has a clinical cutoff of 18 indicating the likely presence of an anxiety disorder (Abramowitz et al. 2010) and has demonstrated good to excellent internal consistency (αs = 0.83–0.96) and convergent validity (rs = 0.54–0.71; Abramowitz et al. 2010). Coefficient alphas for the current sample were 0.87, 0.90, 0.93, and 0.93 for each subscale respectively, and 0.93 for the total score.
Eating disorder symptoms
The Eating Attitudes Test – 26 (EAT-26; Garner and Garfinkel 1979) assessed broad eating pathology with a clinical cutoff of 20 indicating the likely presence of an eating disorder (King 1991). The EAT-26 is highly correlated with the EAT-40 (r = 0.98) and has demonstrated excellent internal consistency (Garner and Garfinkel 1979). Coefficient alpha for the current sample was 0.88.
The Eating Disorder Examination-Questionnaire-6 (EDE-Q-6; Fairburn and Beglin 1994) measured specific eating pathology in the past 28 days. The four subscales (dietary restraint, eating concerns, shape concerns, and weight concerns) demonstrate adequate to excellent internal consistency (αs = 0.78–0.93; Luce and Crowther 1999) and good concurrent validity (rs = 0.68–0.84; Mond et al. 2004). Consistent with past research (Peterson et al. 2007), weight and shape concerns were evaluated together as a single construct, body image concerns. Coefficient alphas for the current sample were 0.85 for dietary restraint, 0.83 for eating concerns, and 0.95 for shape and weight concerns.
The Perceived Stress Scale-4 (Cohen 1988) was used as a measure of general stress to assess the degree to which participants view general situations in their lives (not specific to 2019-nCOV) as stressful in the past 4 weeks (Cohen et al. 1983). The four-item scale has demonstrated acceptable reliability (αs = 0.79) and convergent validity with measures of depression (r = 0.67) among a sample of pregnant women (Karam et al. 2012). Coefficient alpha for the current sample was 0.80.
Four items assessing distress specific to 2019-nCOV were added to the survey on April 3, 2020. Visual analogue scales were used where 0 = not at all and 100 = extremely for the prompts: In the past two weeks… (1) How anxious has COVID-19 made you feel?; (2) How stressed has your life felt due at least in part to COVID-19?; (3) How worried have you been that you or someone you love might get COVID-19?; and (4) In general, how distressed have you felt because of COVID-19? Items were averaged to create a mean score. Only 89 postpartum and 110 control women both completed the survey during the 2019-nCOV pandemic (defined as after March 11, 2020) and received the 2019-nCOV distress items. Coefficient alpha for the current sample was 0.89.
T-tests and chi-square analyses compared postpartum and control women on demographic variables. Correlations between demographic and outcome variables were also performed. Any demographic variable that both differed by group and was significantly correlated with an outcome variable was used as a covariate in analyses for that outcome.
For aim 1, descriptive statistics were calculated to determine the percentage of participants who exceeded clinical cutoffs for depressive, OCD-type, and eating disorder symptoms. Multivariate analysis of variance (MANOVA) models compared participants across the set of eating disorder symptoms (broad eating pathology, dietary restraint, eating concerns, and body image concerns) and the set of general anxiety and OCD-type symptoms. An ANOVA model compared participants on depressive symptoms. For all models, group (coded 0 = control women and 1 = postpartum women) was the independent variable, with the set of eating disorder symptoms, the set of anxiety and OCD-type symptoms, and depressive symptoms evaluated as dependent variables. If covariates were deemed necessary, analyses were multivariate analysis of covariance (MANCOVA) and analysis of covariance (ANCOVA).
For aim 2, in the subsample with data on 2019-nCOV distress (n = 89 postpartum and n = 110 control women), hierarchical linear regression models were evaluated. Mental health history, general stress, and any demographic variables were entered into the first step as covariates, and 2019-nCOV distress and group were entered into the second step. The third step included the interaction between 2019-nCOV distress and group. All models were evaluated separately for dependent variables (i.e., depressive, anxiety and OCD-type, and eating disorder symptoms). Significant interactions were probed using the PROCESS (Hayes 2013) macro for SPSS to analyze simple slopes. For any models in which the two-way interaction was non-significant, main effects were evaluated collapsing across groups.
Given the large number of analyses run, we used the Benjamini–Hochberg procedure to correct for multiple comparisons using a false discovery rate of 0.05 (Benjamini and Hochberg 1995; Thissen et al. 2002). Statistical analyses were conducted using IBM SPSS Statistics version 26.