Behavioral coping phenotypes and psychosocial outcomes in a national U.S. sample of pregnant and postpartum women during the COVID-19 pandemic

Maternal stress exposure during the COVID-19 pandemic may have transgenerational effects, adversely affecting both the pregnant woman and her offspring. Therefore, there is an urgent need to characterize the coping styles and psychosocial distress of pregnant and postpartum women during the COVID-19 pandemic to help mitigate lasting sequalae on both mothers and infants. Here we use latent prole analysis to examine patterns of behavioral coping strategies associated with risk and resiliency to adverse mental and physical health outcomes. Leveraging a large U.S. sample of perinatal women (N = 2,876 pregnant women, N = 1,536 postpartum women), we identied four behavioral phenotypes of coping strategies: (1) passive-coping, characterized by primarily engaging in high levels of screen time, social media use, and eating comfort foods; (2) active-coping, characterized by primarily engaging in high levels of self-care, social support, and limiting media exposure; (3) low-coping, characterized by low levels of all coping strategies; (4) high-coping, characterized by high levels of both active and passive coping strategies. Critically, we found that passive-coping phenotypes were associated with higher levels of depression and anxiety and worsening stress and energy levels in both pregnant and postpartum women. Supplementing passive coping strategies with high levels of active coping strategies (the high-coping prole) lessened adverse outcomes in postpartum women. These behavioral coping phenotypes highlight potential risk and protective factors for perinatal women, which is critical in helping to identify and treat perinatal women most at risk for experiencing mood and affective disorders resulting from the COVID-19 pandemic.


Introduction
The current COVID-19 pandemic re ects a unique, chronic stressor that likely has wide-ranging consequences for psychosocial functioning in pregnant and postpartum women across the globe. Indeed, emerging reports indicate that the COVID-19 pandemic is associated with heightened psychological distress in the general adult population (O'Connor et al., 2020;Xiong et al., 2020), with women and unpaid caregivers reporting disproportionate increases in symptoms of anxiety and depression (Daly, Sutin, & Robinson, 2020). The impact on pregnant and postpartum women is of particular concern given the established adverse effects of perinatal mood and anxiety disorders on the intrauterine and postnatal development of their offspring (Hoffman, Dunn, & Njoroge, 2017;Monk, Lugo-Candelas, & Trumpff, 2019; Van den Bergh et al., 2020). There is an urgent need to characterize the mental health outcomes of perinatal women during the COVID-19 pandemic and to identify behavioral risk and protective factors to minimize potentially harmful consequences during this global public health emergency. Leveraging a large, national sample of pregnant and postpartum women in the United States (N = 4,412), we use a data-driven, person-oriented approach to (1) classify behavioral phenotypes of coping strategies that perinatal women are engaging in to manage pandemic-related stress, (2) isolate associations between coping phenotypes and demographic characteristics, and (3) identify coping phenotypes that are associated with risk and resiliency for adverse mental and physical health outcomes. The ndings from this study are essential for identifying behavioral markers of women at increased risk for adverse outcomes, pinpointing potential protective factors, and targeting interventions to reduce long-term consequences on women and infants.
Pregnant and postpartum women are at heightened risk for mood and anxiety disorders, particularly following stressful life events (Biaggi, Conroy, Pawlby, & Pariante, 2016;Salm Ward, Kanu, & Robb, 2017). The COVID-19 pandemic presents a number of unique stressors that may make perinatal women especially vulnerable to experiencing mood and anxiety disorders. For example, uncertainty regarding the impact of COVID-19 infection or vaccination on fetuses and infants, changes or disruptions in birth plans or postpartum care, and reduced access to childcare or social support systems may create additional risk for maladaptive outcomes in perinatal women relative to the general population. Women are also more likely to work in professions that have increased virus exposure risk, such as healthcare and teaching, and in professions that have experienced greater economic consequences and job losses, such as hospitality or service industries (Zamarro, Perez-Arce, & Prados, 2020). School and childcare closures have also led to a disproportionate increase in unpaid labor taken on by women (Kalenkoski, Pabilonia, & Kalenkoski, 2020;Zamarro et al., 2020).
Collectively, these factors suggest that pregnant and postpartum women might be especially vulnerable to adverse mental and physical health outcomes during the COVID-19 pandemic.
The impact of pandemic-related stressors on pregnant and postpartum women is of heightened concern given that maternal mental health can have intergenerational in uences on their child's development and long-term psychological health (Glover, 2014;Monk et al., 2019). It is well established that anxiety and depression during pregnancy is associated with a number of detrimental outcomes, including increased risk of preterm birth, low birth weight, postpartum depression, and long-term adverse neurobehavioral outcomes in infants (Dunkel Schetter & Tanner, 2012;Glover, 2014;Grote et al., 2010;Monk et al., 2019). Preventing or attenuating the incidence of perinatal mood and anxiety disorders is essential for preventing a sequalae of intergenerational transmission and negative developmental consequences. Yet, the degree to which the COVID-19 pandemic has uniquely impacted perinatal mental and physical health in the United States, as well as the potential behavioral strategies that might promote risk or resilience, has not been well characterized.
Insight into strategies to prevent potentially harmful short-and long-term consequences of pandemic-related stress on perinatal women and their offspring may be gleaned by exploring factors that are associated with risk and resiliency to pandemic-related stressors. For instance, emerging reports indicate that psychological exibility and increased tolerance of uncertainty are associated with resiliency to pandemic-related stress (Daks, Peltz, & Rogge, 2020;McCracken, Badinlou, Buhrman, & Brocki, 2021). Another, large-scale online survey of adults (N = 3,042) found that psychological resiliency, de ned by increased self-reliance, emotion-regulation, and interpersonal relations, was associated with lower COVID-19 related distress (Barzilay et al., 2020). However, behavioral coping strategies associated with risk and resiliency, which are more immediately modi able than psychological characteristics and thus represent potential targets for intervention, have not been well characterized in perinatal women during the COVID-19 pandemic. Filling this empirical gap is critical for helping identify women most at risk for experiencing untreated mood and affective disorders during the pandemic and for informing scalable interventions to help mitigate lasting sequalae on women and their infants.
Here we use a data-driven, person-oriented approach to identify behavioral phenotypes of coping mechanisms in a large, national sample (N = 4,412) of pregnant and postpartum women (within the rst 12 months of infant life) drawn from nine states across the United States. While traditional variable-oriented approaches assume that the associations among variables within a population are homogenous, person-oriented approaches, such as latent pro le analysis, identify heterogenous subgroups that have similarities across several variables of interest. This approach is advantageous as it considers how patterns of behaviors jointly predict outcome measures, which allows for more precise identi cation of behavioral phenotypes associated with risk and resiliency.
As such, we rst use latent pro le analysis to identify different adaptive and maladaptive behavioral coping phenotypes in pregnant and postpartum women. We examined behavioral phenotypes in pregnant and postpartum women separately to account for potentially different patterns of coping strategies, and relations with outcome measures, that might be obscured in the full sample (e.g., increased substance use may be more common among postpartum women relative to pregnant women). We then examine relations between coping phenotypes and demographic/socioeconomic characteristics of our sample. Finally, we identify coping phenotypes that are associated with risk and resiliency to maternal anxiety and depression symptoms, as well as measures of sleep quality, stress and energy levels, and overall perceived distress related to the pandemic. Leveraging this large, geographically diverse sample, the current study characterizes the impact of the COVID-19 pandemic on perinatal psychosocial functioning and offers insights into potential risk and protective factors for adverse outcomes.

Method
Participants & procedures Participants were recruited into studies examining the impact of COVID-19 on perinatal women taking place at 14 academic research institutions ( Figure 1A). All studies were approved by the local Institutional Review Board at each site.
The studies were independent, but the investigators opted to use common research methods to facilitate future harmonization and cross-site data sharing. All sites administered the COPE: COVID-19 & Perinatal Experiences -Impact Survey (Thomason, Graham, & VanTieghem, 2020) online between March and October 2020, with the majority of data collection occurring in April 2020 ( Figure 1B). Informed electronic consent was obtained prior to data collection. Criteria for participation included being pregnant or postpartum within the rst 12 months of infant life. The combined sample consisted of 4,412 women (2,876 pregnant women, 1,536 postpartum women) collected from 9 states (California, Michigan, New York, Oregon, Pennsylvania, Tennessee, Utah, Vermont, Virginia; Figure 1). Demographic characteristics of the nal sample are reported in Table 1. The survey was administered to an additional 1,294 women, but their data were excluded due to failure to complete the "Adjustment and Coping" subsection of the COPE-IS survey (n = 814), or due to inconsistent or unreliable survey response patterns (n = 37 were excluded due to selecting categories for both "increased" and "decreased" behaviors; n = 389 were excluded due to selecting "None" in addition to other behaviors). We additionally excluded n = 54 women who reported no coping behaviors.

Measures
Coping behaviors. Women selected behaviors they were engaging in to cope with stress related to the COVID-19 pandemic on the "Coping and Adjustment" subsection of the COPE: COVID-19 & Perinatal Experiences -Impact Survey (Thomason et al., 2020). The speci c survey items and proportion of subjects endorsing each item are listed in Figure 2. The survey items probed a broad spectrum of adaptive and maladaptive behaviors, including social support mechanisms, physical activity, nutrition, substance use, healthcare utilization, news consumption, screen time, and selfcare related activities ( Figure 2). One additional item ("using other recreational drugs") in the original survey was removed for analyses due to fewer than 5 responses to this question across participants. The survey items were selected to encompass a range of physical, emotional, and social support mechanisms that are both accessible to women during the pandemic and that are hypothesized to have predictive value for risk and resiliency for poor mental and physical health outcomes.
Mental and physical health outcomes. Data were collected on how the COVID-19 pandemic changed women's self-reported sleep, energy levels, and overall stress/mental health using 5-point Likert scales (1 = worsened signi cantly, 2 = worsened moderately, 3 = no change, 4 = improved moderately, 5 = improved signi cantly). We also collected data on women's self-reported overall level of stress speci cally related to the COVID-19 pandemic using a 7-point Likert scale (1 = nothing, 7 = extreme). Finally, women's overall levels of anxiety and depression symptoms were measured using the Anxiety and Depression subscales of the Brief Symptom Inventory (BSI-18; Derogatis, 2001), as well as clinically-elevated symptoms of depression and anxiety, de ned by BSI transformed t scores of 63 or greater (Derogatis, 2001).

Analytic plan
We rst used factor analysis to reduce the survey items into composite variables representing different dimensions of coping strategies. We conducted an exploratory principal components analysis to determine the latent structure of coping strategies using SPSS version 21.0, which was veri ed using con rmatory factor analysis in Mplus Version 8.1.
We then used latent pro le analysis (LPA) to identify groups of pregnant and postpartum women categorized based on similar patterns of coping strategies. All LPA analyses were conducted in Mplus Version 8.1, and separate models were t for pregnant women and postpartum women to account for potential differences in patterns of coping strategies between pregnant and postpartum women. Each LCA was initialized 200 times, with 50 iterations for the nal stage of optimization. Using the Nylund, Asparouhov, & Muthén (2007) guidelines for latent pro le analysis, the following steps were used to determine the best tting model. As a rst step, the model with the lowest Bayesian information criterion (BIC) combined with a statistically signi cant Lo-Mendell-Rubin likelihood ratio test (LMR) was considered as the potentially best tting model. As a second step, we ensured that the best tting model also had a high entropy value (greater than .75, or closest to 1.0, indicating low classi cation error), at least 5% of the total participant count in a given pro le, and high posterior probabilities (close to 1.0, indicating high con dence that an individual assigned to a given pro le actually belongs to that pro le). We sequentially tested the t of multiple models, stopping once the LMR indicated no difference.
Finally, we examined demographic and socioeconomic predictors of latent pro le membership using the 3-step procedure for predictor variables ("R3STEP" command for multinomial logistic regression in Mplus). To safeguard against biased estimates when examining associations between pro le membership and distal outcomes (stress levels, sleep quality, energy levels, symptoms of depression and anxiety, and overall COVID-related distress), we used the 3-step auxiliary approach for outcome measures with unequal means and equal variances ("DE3STEP" command in Mplus).
This approach accounts for measurement error associated with most likely pro le membership and is shown to be superior to other methods (Asparouhov & Muthén, 2014).

Descriptive statistics
Demographic characteristics and means for all mental and physical health outcome variables for the full sample are presented in Table 1. The proportion of participants endorsing each of the coping behavior survey items are presented in Figure 2. Twenty-one percent (21.2%) of pregnant women and 20.8% of postpartum women met cut-off estimates for clinically elevated anxiety symptoms, and 16.6% of pregnant women and 16.3% of postpartum women met cut-off estimates for clinically elevated depression symptoms (both de ned by BSI transformed t scores of 63 or greater; Derogatis, 2001 a Education was coded as 1 = < 10 th grade, 2 = 10-12 th grade, 3 = high school/GED, 4 = apprenticeship/trade school, 5 = partial college, 6 = 2-year college, 7 = 4-year college, 8 = graduate degree b Income was coded as 1 = < 10k, 2 = 10-20k, 3 = 20-30k, 4 = 30-40k, 5 = 40-50k, 6 = 50-60k, 7 = 60-80k, 8 = 80-100k, 9 = 100-120k, 10 = 120-140k, 11 = 140-160k, 12 = 160-180k, 13 = 180-200k, 14 = 200-220k, 15 = 220-250k, 16 = 250k+ Dimensionality reduction Exploratory factor analysis. We used a principal components analysis with promax rotation, which permits correlation between the factors, to guide the creation of composite variables from the original survey items (Figure 2). Examination of the scree plot suggested a 6-factor solution, which was conceptually appropriate and accounted for 38.7% of the overall variance. One item ("Other") did not load on any factors above .2 and was thus removed. All remaining variables had loadings above .45, with the exception of "Using CBD only", which showed a marginally lower loading of .37. There were no substantial cross-loadings of variables onto multiple factors. The structure matrix for the nal 6-factor solution is shown in Table 2. Factor 1 re ects "self-care" and consists of exercising, getting a good night's sleep, meditation, eating healthy, self-care activities (e.g., baths, facials), and calm activities such as puzzles and reading. Factor 2 re ects "vegging out" and consists of increased screen time, increased social media use, and increased comfort foods. Factor 3 re ects "decreased media/news" and consists of decreased social media use, decreased time following news, and increased time following news (reverse scored). Factor 4 embodies "social support" and consists of talking with friends and family, helping others, engaging in family activities, and talking to other parents/pregnant women. Factor 5 re ects "healthcare utilization" and consists of talking to health providers, talking to mental health providers, using new prescription drugs, and using overthe-counter sleep aids. Finally, Factor 6 re ects "substance use" and consists of tobacco use, marijuana, CBD, and alcohol consumption. Correlations among factors ranged from -.127 to .195. Broadly, these factors re ect dimensions of active coping strategies (self-care, social support, healthcare) and passive coping strategies (vegging out, substance use, and media use).
Con rmatory factor analysis. We veri ed the t of these active and passive coping strategy dimensions using con rmatory factor analyses in Mplus with Robust Weighted Least Squares (WLSVM) extraction.  (Kline, 2015). Inspection of parameter estimates indicated that all of the factor loadings were signi cant and in the expected direction for both models. Composite reliability (CR), using McDonald's omega, indicated that internal consistency was adequate for self-care (CR = .70), vegging out (CR = .74), decreased media/news (CR = .89), substance use (CR = .75), and healthcare utilization (CR = .72), based on recommended guidelines of composite reliability equal to or greater than .60 for exploratory scales (Hair, Black, Babin, & Anderson, 2010). Social support had relatively lower internal consistency (CR = .55), but we retained this composite given the conceptual t and given that all indicators loaded signi cantly and in the expected direction on this factor.
Coping strategy composite variables. Composite variables for "self-care", "social support", "decreased media/news", and "vegging out" were created by averaging over the individual items that loaded strongly onto each of these constructs ( Table 2). Note that "increased news coverage" was reverse coded prior to averaging in the "decreased media" composite variable. Composite variables for "substance use" and "healthcare utilization" were created by discretizing the individual items into a binary categorical variable (none, or 1+ items). We did this to prevent oor effects resulting from the low proportion of responses across these items (see Figure 2), which would have prevented model convergence in subsequent latent pro le analyses. We created composite variables in this way, rather than using the latent factor scores, as this method retains the variance in the original data and is superior when using exploratory scales (DiStefano, Zhu, & Mîndrila, 2009). Preserving the variation in the original data is particularly desirable here as the factor structure explained only 39% of the total variance.

Latent pro le analysis
Four models, testing the t of 2-5 possible pro les (stopping once the LMR test was no longer signi cant), were generated and compared for both pregnant and postpartum women separately using the composite coping strategy variables as indicators. The 4-pro le model was the best tting model for both pregnant and postpartum women, as re ected by a low BIC, a signi cant LMR test, a high entropy value, and each pro le representing at least 5% of the entire sample (see Table 3 for full results). The latent structures of the pro les are presented in Figure 3, which illustrate mean values for the continuous and categorical variables for each pro le. The means of the coping strategies were directly compared between pro les using an analysis of variance (ANOVA), which revealed signi cant differences among the pro les on the behavioral coping strategies (see Table 4 and Table 5). Tukey's Honest Signi cant Difference test was used to compare the means of the individual coping strategies between the pro les (Table 4 and Table 5). Similar pro les were found for pregnant and postpartum women and are described below.   Differing subscripts within rows indicate signi cantly different means at p < .05, with Tukey's HSD correction.
Pro le 1 -Low-coping. The rst pro le extracted accounted for 41% of the sample in pregnant women (n = 1188), and 39% of the sample in postpartum women (n = 595). This pro le was characterized by signi cantly lower endorsements of all coping strategies in both postpartum and pregnant women.
Pro le 2 -Passive-coping. The second pro le extracted accounted for 33% of the sample in pregnant women (n = 960), and 41% of the sample in postpartum women (n = 635). This pro le was characterized by high levels of vegging out, and lower levels of self-care, social support, and healthcare utilization.
Pro le 3 -Active-coping. The third pro le extracted accounted for 12% of the sample in both pregnant women (n = 349) and in postpartum women (n = 178). This pro le was characterized by high levels of self-care and social support, as well as decreased media/news consumption.
Pro le 4 -High-coping. The nal pro le extracted accounted for 13% of the sample in pregnant women (n = 379), and 8% of the sample in postpartum women (n = 128). This pro le was characterized by high levels of self-care, social support, and healthcare utilization, as well as very high levels of vegging out. This pro le was also characterized by higher levels of substance use in the postpartum women only.

Relation to demographic and socioeconomic variables
Multinomial logistic regression analysis was used to examine predictors of latent pro le membership based on demographic and socioeconomic variables. The multinomial logistic model parameters using the low-coping pro le as the reference category are presented for pregnant and postpartum women in Table 6 and Table 7, respectively. Pregnant women in the active-coping pro le were more likely to have fewer children (odds ratio [OR] = .64, 95% CI = .52-.78).
Women in the passive-coping pro le were also more likely to identify as Black (OR = 2.33, 95% CI = 1.42-3.84). Women in the active-coping and high-coping coping pro les were more likely to have greater educational attainment (active-coping: OR = 1.26, 95% CI = 1.07-1.48; high-coping: OR = 1.25, 95% CI = 1.07-1.47). No other variables were signi cant predictors of pro le membership. Relation to mental and physical health outcomes Finally, we examined whether latent pro le membership was a signi cant predictor of mental and physical health outcome variables. In pregnant women, membership in a particular pro le was a signi cant predictor of all outcome measures, as indicated in Table 8. Of note, women with coping pro les characterized by high levels of vegging out (highcoping and passive-coping) were more likely to have increased anxiety and depression ( Figure 4). Women in these pro les also exhibited worsening stress levels, and overall greater COVID-related distress relative to women in the activecoping and low-coping pro les. Additionally, women in the active-coping pro le reported fewer changes in energy levels relative to the other pro les. Women with pro les characterized by high self-care and social support (active-coping and high-coping) also had fewer negative changes in sleep quality. As indicated in Table 9, latent pro le membership was a signi cant predictor of all outcome variables for postpartum women with the exception of changes in sleep quality, which showed no differences. Of note, women in the passivecoping pro le reported elevated depression symptoms relative to all other pro les ( Figure 4). Additionally, women in the passive-coping pro le also exhibited worsening stress levels relative to other women. We also found that women with pro les characterized by high levels of vegging out (high-coping and passive-coping) reported increased anxiety symptoms (Figure 4), as well as moderately worsening energy levels. [1] We veri ed that model t did not change if conducting the con rmatory factor analyses on pregnant and postpartum women separately.

Discussion
There is urgent need to characterize the impact of COVID-19 related stress on perinatal women, particularly given the potential for signi cant stressors to have negative intergenerational effects on their offspring (Monk et al., 2019;Van den Bergh et al., 2020). In addition to presenting a number of unique health, economic, and social-related stressors, COVID-19 has also disrupted routines, support systems, and behavioral strategies that previously may have been bene cial for mitigating psychological distress in response to stressful life events. Identifying behavioral phenotypes that are related to risk and resiliency for negative outcomes is essential for targeting resources and interventions for those most vulnerable. Here we used person-oriented latent pro le analyses to identify behavioral phenotypes of coping strategies in a large, national sample of perinatal women. Using a person-oriented approach, over a variable-oriented approach, provides a better understanding of how distinct patterns of behavioral coping strategies jointly predict outcome measures. This approach thus affords more precise identi cation of potential risk and protective factors for mental and physical health outcomes.
Across both pregnant and postpartum women, we observed four distinct phenotypes of coping strategies. The most common phenotype (low-coping pro le) consisted of women who reported low levels across all categories of coping strategies. The second most prevalent phenotype (passive-coping pro le) consisted of pregnant and postpartum women who primarily engaged in passive coping strategies characterized by "vegging out" -that is, engaging in increased screen time, social media use, and comfort foods to cope with pandemic-related stress. We also identi ed phenotypes of women who reported high levels of all coping strategies (high-coping pro le), and women who predominately reported engaging in increased self-care, social support, and limiting exposure to news coverage and social media (active-coping pro le). Additionally, women in the high-coping phenotype reported increased substance use, however this was only true for postpartum women. Pregnant and postpartum women in both the high-coping and active-coping phenotypes tended to report higher levels of education, and pregnant women in these phenotypes also had fewer children than other women.
We also examined whether pro le membership differed by race/ethnicity, particularly given that Black, Indigenous, and people of color (BIPOC) have been disproportionally impacted by the COVID-19 pandemic (Cyrus et al., 2020;Millett et al., 2020). In pregnant women, we observed that Black women were 35% less likely to be members of the passive-coping phenotype, and 70% less likely to be members of the active-coping phenotype. We also found that Asian women were 58% less likely to be members of the active-coping phenotype. In postpartum women, Black women were twice as likely to be members of the passive-coping phenotype, but we observed no other differences based on race/ethnicity. To ensure that pro le identi cation was not skewed by race/ethnicity, we conducted post hoc latent pro le analyses with only BIPOC women, which indicated the same patterns of behavioral pro les and associations with outcome measures (see Supplementary Information). These ndings suggest that similar patterns of behavioral coping phenotypes are observed regardless of race/ethnicity, but the probability of belonging in a particular coping phenotype varies by race/ethnicity. This is particularly noteworthy given differential associations between pro le membership and mental and physical health variables. However, these ndings should be interpreted with caution given that our sample was somewhat skewed, with 71% of the sample identifying as non-Hispanic white (see Table 1).
A key aim of this study was to identify associations between coping phenotypes and measures of mental and physical health. Both pregnant and postpartum women reported worsening changes in sleep quality, energy levels, and stress levels relative to pre-pandemic levels, but the degree to which women reported worsening outcomes differed based on coping phenotype. In particular, women in phenotypes characterized by high levels of vegging out (the high-coping and passive-coping pro les) reported the largest negative change in energy and stress levels. In contrast, women in the active-coping phenotype reported fewer negative changes in sleep quality and energy levels relative to other women, particularly among pregnant women. Similarly, women in the low-coping phenotype also reported fewer negative changes in mental and physical health outcomes, as well as relatively lower levels of overall COVID-19 related distress.
This suggests that women in the low-coping phenotype may re ect a subgroup of women who are experiencing lower distress related to the pandemic and are accordingly engaging in fewer coping strategies.
When examining mental health outcomes, we observed high levels of clinically elevated symptoms of both anxiety (21.2% of pregnant women and 20.8% of postpartum women) and depression (16.6% of pregnant women and 16.3% of postpartum women) across the full sample. For reference, recent meta-analyses estimate that pre-pandemic rates of perinatal anxiety and depression in the United States were approximately 20.7% and 11.9%, respectively (Fawcett, Fairbrother, Cox, White, & Fawcett, 2019;Woody, Ferrari, Siskind, Whiteford, & Harris, 2017). This suggests that rates of perinatal depression in particular may be elevated relative to pre-pandemic levels. Notably, we found signi cant differences in reported levels of anxiety and depression symptoms as a function of behavioral coping phenotypes.
Pregnant women in the passive-coping and high-coping phenotypes reported increased symptoms of both anxiety and depression (Figure 4). Similarly, postpartum women in passive-coping and high-coping phenotypes reported increased anxiety symptoms; however postpartum women in the high-coping phenotype exhibited fewer depression symptoms relative to women in the passive-coping phenotype (Figure 4). This result suggests that supplementing passive coping strategies with more active strategies may buffer the negative association between passive coping and increased depression symptoms, particularly among postpartum women. Yet, supplementing passive coping strategies with active coping strategies did not lessen the severity of anxiety symptoms in either pregnant or postpartum women.
While we cannot establish causality, our ndings are consistent with clinical models of depression, which emphasize the role of inactivity as both a symptom of and contributor to a self-reinforcing cycle of depression (Martell, Dimidjian, & Herman-Dunn, 2013). Moreover, our ndings align with recent reports indicating that increased physical activity is associated with increased resiliency to COVID-19-related stress in pregnant women across the globe (Davenport, Meyer, Meah, Strynadka, & Khurana, 2020;Lebel, MacKinnon, Bagshawe, Tomfohr-Madsen, & Giesbrecht, 2020;Preis, Mahaffey, Heiselman, & Lobel, 2020). However, even though we observed moderately elevated levels of depression relative to prepandemic estimates, we cannot ascertain that our ndings are directly tied to the COVID-19 pandemic, particularly given the lack of a matched pre-pandemic comparison group. Indeed, these behavioral phenotypes may not be unique to the COVID-19 pandemic but may re ect more generalizable patterns of coping strategies in response to stressful life events.
For instance, prior work has similarly observed that engaging in passive psychological coping styles, such as avoidance or denial, is associated with increased depression symptoms during the prenatal and postpartum periods (Gutiérrez-Zotes et al., 2016;Honey, Bennett, & Morgan, 2003;Razurel, Kaiser, Sellenet, & Epiney, 2013;Van Bussel, Spitz, & Demyttenaere, 2009). Yet, it is noteworthy that we observed a large percentage of both pregnant and postpartum women in the passive-coping behavioral phenotypes in our sample (33% and 41%, respectively). Future work is needed to ascertain whether these rates are substantially higher than might typically be expected.
In pregnant women, supplementing high levels of passive coping strategies with active strategies (i.e., the high-coping pro le) did not substantially lessen the severity of depression symptoms. This result is in contrast to our ndings in postpartum women, where we observed attenuated depression symptoms among women in the high-coping pro le. This is an important consideration when evaluating both the etiology of symptoms and recommendations for informing perinatal depression interventions. One explanation for this result could relate to differences in pregnant versus postpartum women's motivation for engaging in active coping strategies. For instance, pregnant women may engage in active strategies such as increased physical activity and healthy eating for the wellbeing of their unborn baby, whereas postpartum women may be more likely to engage in these activities for their own health and wellbeing. Another possibility could be due to relatively increased vulnerability of pregnant women to mood and anxiety disorders (Kessler, 2003;Underwood, Waldie, D'Souza, Peterson, & Morton, 2016), which may negate the otherwise bene cial effects of behavioral activation for depression (Martell et al., 2013). Nonetheless, even though we cannot determine the source of these differences, these ndings demonstrate that "vegging out" may be an important behavioral marker for adverse outcomes in pregnant women. This knowledge can be used to target interventions and support systems for women at risk for perinatal mood and anxiety disorders, possibly even prior to the emergence of clinically elevated symptoms. For instance, clinicians and care providers could have women use wearable activity monitors or ask questions probing physical activity and nutrition to help identify and provide resources for women at risk for adverse outcomes. This information is especially relevant given that a number of women often do not report symptoms or seek treatment for perinatal mood and anxiety disorders, in part due to stigma surrounding maternal depression or fear of teratogenic effects with medication use (Bonari et al., 2004).
The COVID-19 pandemic is unique relative to other stressful life events that have impacted widespread communities, such as natural disasters, partly due to the chronic and uncertain nature of the pandemic. Stress related to high levels of uncertainty is proposed to be energetically costly (Peters, McEwen, & Friston, 2017). Thus, individuals may be more likely to "veg out" and eat high-calorie comfort foods as a way to cope with the increased energetic demand associated with uncertainty stress during the COVID-19 pandemic. Indeed, we observed that a large proportion of women were members of passive-coping pro les (33% of pregnant women and 41% of postpartum women). The prevalence of these coping phenotypes might relate to the high energetic cost of uncertainty stress resulting from the COVID-19 pandemic, which could lead more women to "veg out". This idea is additionally supported by recent ndings showing that individual differences in tolerance of uncertainty was related to greater psychological distress at the start of COVID-19 lockdowns in adults (Rettie & Daniels, 2020). Future work is needed to assess this possibility and the long-term physical and psychological costs associated with passive coping strategies to manage pandemic-related stress.
In sum, this large-scale, national study highlights the impact of the COVID-19 pandemic on mental and physical health outcomes of perinatal women in the United States. Importantly, it identi es widespread heterogenous patterns of coping strategies that perinatal women are engaging in to manage pandemic-related stress. This study also illustrates the advantages of using person-oriented approaches to ascertain the complex and multifaceted patterns of behaviors that are associated with differing mental and physical health outcomes. The behavioral coping phenotypes we identi ed highlight potential risk and protective factors for perinatal women, which is critical in helping to identify and treat women most at risk for experiencing mood and anxiety disorders during this global health crisis.

Declarations
Con ict of Interest Statement: The authors report no con icts of interest.

Data Availability Statement:
The data that support the ndings of this study are available from the corresponding author upon reasonable request.

Figure 1
Geographic distribution and study site locations (A), and density plots illustrating the temporal distributions of data collection by state (B).

Figure 2
Percentage of pregnant and postpartum women endorsing each survey item.

Figure 3
Estimated means for the 6 coping strategies across all pro les for both pregnant women and postpartum women.