COVID-19 Pandemic Experiences and Symptoms of Pandemic-Associated Traumatic Stress Among Mothers in the US

This cohort study evaluates the association between COVID-19 pandemic experiences and symptoms of pandemic-associated traumatic stress among mothers in the US.


eTable 1. Description of ECHO Cohorts Included in Analysis
Names N Description ECHO in Puerto Rico (PROTECT) 16 Cohort consists of pregnant women recruited from 5 OB/GYN clinics in Puerto Rico and their offspring. Aim of study is To understand environmental exposures and their effects on birth outcomes and children's neurodevelopmental health in Puerto Rico. Prematurity and Respiratory Outcomes Program (DINE-PROP) 75 Cohort consists of preterm infants ≤7 days old and gestational age (GA) between 23-<29 weeks. Aim of study is to identify clinical, physiologic, and biochemical markers during NICU hospitalization that can predict respiratory morbidity through 1 year of age. Trial of Late Surfactant (DINE-TOLSURF) 37 Cohort consists of preterm infants ≤ 28 weeks GA intubated and mechanically ventilated at 7-14 days with plan to treat with inhaled nitric oxide. Aim of the study is to examine the effect of late doses of surfactant on bronchopulmonary dysplasia (BPD) at 36 weeks postmenstrual age (PMA) in preterm infants receiving inhaled nitric oxide. NICU Hospital Exposures and Long-Term Health (DINE-NICUHEALTH) 48 Cohort consists of very low birth weight infants recruited upon admission to the NICU at a New York City hospital and their caregivers. Aim of study is to identify and assess the impact of NICU-based chemical and non-chemical environmental exposures on neurobehavioral outcomes through early childhood Preterm Erythropoietin Neuroprotection Trial (DINE-PENUT) 27 Cohort population consists of infants 24 hours of age or less and between 24-<28 weeks GA admitted to the NICU. Aim of this study is to test the efficacy of erythropoietin (Epo) for the neuroprotection of extremely premature infants and assess whether early high dose Epo improves survival without neurodevelopmental impairment. Family Life Project (FLP) 473 Cohort consists of population-based primary caregiver-child dyads (and secondary caregiver when available), with over-sampling for poverty and African American participants, in predominantly lowincome, non-urban counties recruited at birth from 1 of 7 hospitals in Pennsylvania or North Carolina. Aim of this study is to investigate associations between psychosocial early life stress and neurodevelopment, including executive function, emotion regulation, language development, school achievement, and attention deficit/hyperactivity disorder (ADHD). 43rd Multicenter Airway Research Collaboration (MARC-43) 82 Serves as a cohort of healthy "controls" for the MARC-35 cohort, an ongoing, 17-center cohort comprised of 1016 infants initially hospitalized for bronchiolitis, of whom 921 are in long-term followup. Comprised of two groups: a) 120 healthy infants who were enrolled between November 2013 through May 2014 from a primary care group practice at Massachusetts General Hospital (MGH); and b) an additional 600 healthy infants from four U.S. sites that participated in MARC-35. Primary outcomes are recurrent wheezing by age 3 years, physician-diagnosis of asthma by age 5 years, and asthma diagnosis and phenotypes at 6 years by exam.

Healthy Start 184
Cohort consists of Colorado community sample of mother-child dyads recruited in infancy from obstetrics clinics at a university hospital and by word of mouth, as well as medical university employees. Aim is to understand the contribution of metabolic and behavioral factors during pregnancy to the development of childhood obesity, insulin resistance, and inflammatory markers.

BAMBAM 103
Cohort consists of Children recruited between 3 months to 12 years of age from the general population recruited in pediatrician offices and via online webpages, radio advertisements, and flyers in Providence, RI. Aims are to examine typical brain development, including how brain growth is altered by specific pre-and post-natal environmental or genetic factors; how patterns of brain growth are associated with, and predictive of, emerging cognitive and behavioral abilities; and how these brain-behavior relationships are influenced by modifiable factors experienced throughout childhood MINNIE 87 Cohort consists of Children recruited between 3 months to 12 years of age from the general population recruited in pediatrician offices and via online webpages, radio advertisements, and flyers in Providence, RI. Aims are to examine typical brain development, including how brain growth is altered by specific pre-and post-natal environmental or genetic factors; how patterns of brain growth are associated with, and predictive of, emerging cognitive and behavioral abilities; and how these brain-behavior relationships are influenced by modifiable factors experienced throughout childhood Cohort consists of biological parents, adoptive parents, and their children recruited at the time of birth through adoption agencies in Oregon, Washington, Texas, Maryland, Pennsylvania, Virginia, New Jersey, California, Utah, Minnesota, Illinois, and Florida. The primary study aims are to examine how family, peer and contextual processes affect children's adjustment, and to examine their interplay (mediation, moderation) with genetic influences. Early Parenting of Children (EPoCh)

33
Sample consists of 200 sibling pairs, recruited from one site in Oregon, in which one sibling was reared from birth with an adoptive family, and the other sibling was reared from birth by the biological mother of the sibling pair. Utilizes a naturalistic human crossfostering design to advance the understanding of childhood pathways to risk for drug use by identifying nuances in the rearing environment that are associated with child risk behaviors and competencies. Navajo Birth Cohort Study (NBCS) / Understanding Risk Gradients from Environment on Native American Child Health 160 Sample is comprised of Native American mothers who have lived on Navajo Nation or within adjacent community for at least 5 years and their children, recruited from sites in New Mexico and Arizona. Aims to evaluate the influence of prenatal and postnatal exposure to metals on birth, neurodevelopmental, immunological, and obesity outcomes.
Cohort consists of children of women recruited prenatally from 3 OB/GYN sites in Boston, MA, St. Louis, MO, and San Diego, CA whose biological mother or father had a history of asthma, eczema, and/or allergic rhinitis. Aim is to examine the role of allergy, the environment, and genetics in developing asthma as well as investigating whether vitamin D levels in pregnancy prevent or reduce asthma, wheezing, or other allergic illnesses in childhood. Cohort consists of pregnant women with a biological child with autism spectrum disorder (ASD) and the offspring of the current pregnancy, recruited from local community events, autism conferences, direct mail via autism service providers, and online and social media. Aim is to understand pre-, peri-, and postnatal ASD risk factors and risk biomarkers. Cohort consists of pregnant women recruited from 2 hospitals in San Francisco, CA and their offspring and aims to examine the unique and combined impact of prenatal exposure to endocrine disrupting chemicals and maternal stress on child neurodevelopmental health and functioning. National Children's Study -IVS Salt Lake County (NCS-IVS-SLC) + Associate

64
A population-based sample of pregnant women, or women planning pregnancy, enrolled from Salt Lake County, Utah that examines the environmental influences on children's health and development. The "environment" was defined broadly to include physical environment (neighborhoods and communities, housing radiation); biological (infectious agents, glucose metabolism, diet) and chemical factors (pesticides, air pollution, pharmaceuticals); geography; psychosocial environment (social networks, socioeconomic status, education, behavioral, family, and cultural influences); and genetics (gene-environment and gene-gene interactions). Individual items were summarized as: (1) A total sum score (range: 5 to 50) that reflects general stress severity; (2) A total significant symptoms score, in which each symptom item was re-scored as "1" if participants "sometimes," "often" or "very often" reported the symptom (item score ≥3) or "0" if a symptom was "rarely" or "not at all" experienced (item score <3). These were summed to create a total number of clinically relevant symptoms (range: 0 to 10), reflecting a dimensional measure of traumatic stress; and (3) A total symptom categories score, which reflected the number of symptom categories endorsed at a clinically significant level (range 0-5), e.g., sum of symptoms domains in which at least one within the domain had a score ≥3. The total significant symptoms score is used as the primary outcome in the current study, because it captures the range of pandemic-related traumatic stress across the sample, whereas the symptom categories score reflects only those with clinically significant symptoms in each category. Therefore, the total sum score and symptom categories score outcomes are included in the Online Supplement as secondary outcomes (Supplemental Tables 1, 3, 4).
Pandemic-related hardships and/or daily life impacts. A total of 13 dichotomous variables were examined including the mother or household member testing positive for SARS-CoV-2 (2 items); changes to healthcare (1 item); increased social isolation (5 items: quarantining from family, reduced contact with family, friends, and colleagues, and attending fewer community events), and changes to the mother's or partner's work (5 items: job loss or reduction of hours, working remotely, getting a new job or more hours, laying off employees, or work environment placing mothers/partners at increased COVID-19 risk).
Adaptive/maladaptive coping strategies and behavior/lifestyle changes. A total of 11 coping strategies and 7 behavior/lifestyle changes-with responses of yes/no-were included.
Mothers reported whether they engaged in any of the following coping strategies during the They also reported on behavior/lifestyle changes during the COVID-19 pandemic compared to pre-pandemic (more/less home-cooking, takeout, exercise and time outdoors; or no behavior changes.

Statistical Analysis
All analyses were conducted using R statistical software 3.6.2. K-means clustering was conducted using the packages "cluster", "fpc", and "factoextra". Sparse Partial Least Squares regression was conducted using the package "spls".

K-means clustering
In this cross-sectional study, we characterized clusters of mothers based on their responses to 31 pandemic hardships, coping strategies, and lifestyle/behavior changes using kmeans clustering, 4,5 an unsupervised machine learning method that groups individuals based on similarities in patterns of responses to specified inputs. In this case, the 31 inputs are responses to the COVID-19 questionnaires (pandemic-related hardships, coping strategies, and behavior changes), independent of any outcome of interest. The optimal number of clusters was selected based on scores from a Silhouette analysis, 6,7 which estimates the similarity of data points within clusters versus between clusters; the number of clusters that maximizes the silhouette analysis score is the optimal number of clusters.
We then described the sociodemographic characteristics of these clusters to determine if there were fundamental differences in characteristics by cluster with respect to maternal race, ethnicity, marital status, educational attainment, and/or household income. SPLS is a supervised machine learning approach that 1) allows inclusion of all potential variables into a single regression model and 2) conducts variable selection and shrinkage based on whether or not inputs are determined to be associated with the outcome, relative to the other inputs in the model. The optimal number of variable groupings and penalty for primary analyses was determined with K-fold cross validation, determined based on the values that minimize mean squared prediction error of the model. This method results in the exclusion of some variables from the final summary tables (noted as "dropped" in regression tables). Bootstrapping was used to calculate 95% confidence intervals. We further determined whether the association between responses to pandemic-related hardships and pandemic-related traumatic stress differed by cluster membership by running stratified, cluster-specific SPLS models. Clusters are fully independent of one another in that an individual cannot be a member of both clusters.
Therefore, we used the approach described by Altman. 9 Briefly, the difference between clusters is calculated as the absolute value of the difference in the coefficients, or abs(b1 -b2), where b1 is the estimated coefficient for a hardship in cluster 1 and b2 is the estimates coefficient for that same hardship in cluster 2. The standard error is approximated as the square root of the sum of squared standard errors from each cluster, or SE = sqrt(SE1^2 + SE2^2). Because PLS provides bootstrapped confidence intervals, we calculated the SE1 and SE2 from the bounds of the 95% confidence intervals. The estimates SE was the average of the SE obtained from the upper and lower bounds of the bootstrapped 95% confidence intervals.