Between May and June 2020, we conducted the Capturing Women’s Experiences in Outbreak and Pandemic Environments (COPE) Study, a web-based cross-sectional study of COVID-19 outcomes, testing, and prevention behaviors among adult women in the United States. Our study included participants who were 18 years or older, cisgender or transgender female, living in the United States, and able to understand English. We excluded participants who did not reside within the United States, were younger than 18 years, were male, or did not understand English.

Recruitment occurred through Facebook advertisements, designed to recruit adult women currently residing in the United States. We aimed to collect at least 400 survey responses. Interested women accessed a digital consent form within the REDCap (Research Electronic Data Capture) survey platform. The consent form included the purpose of the study and participant risk, burden, rights, and compensation. Participants who provided documented consent were directed to a screening survey. Screening questions included age, gender, and state of residence. Of 682 users who clicked on an advertisement, 633 (92.8%) provided informed consent and answered screener questions. Among the 633 participants, 7 participants were not eligible for the study. Eligible participants (n=626) were directed to the survey, which took 15 to 20 minutes to complete. Of 626 eligible participants, 491 (78% response rate) completed the full web-based survey. Participants received a US $20 Amazon e-gift card via email. The COPE Study procedures were approved by the University of California, San Diego Human Research Protections Program Institutional Review Board.

Excluding participants without reported race/ethnicity, the sample was 473 women. We computed descriptive statistics of all variables overall and by race/ethnicity. We report medians and IQRs for nonnormally distributed continuous variables and frequencies and proportions for categorical variables. For race/ethnicity comparisons, we used Kruskal-Wallis tests and chi-square tests. If subgroups had expected values less than 5, Fisher exact test or a Monte Carlo estimate for Fisher exact P value was used. Individual pairwise comparisons by race/ethnicity were conducted when the Kruskal-Wallis test or chi-square test had a P<.05. The Dwass-Steel-Critchlow-Fligner test was used when the Kruskal-Wallis test was significant (P<.05), and individual chi-square tests between each of the racial/ethnic groups were examined when the overall chi-square was significant (P<.05). To account for potential confounding, COVID-19–related outcomes, stressors, and prevention behaviors that were significantly associated in bivariate analyses were included in multivariable binary logistic regressions. Each regression was adjusted for age, education, income, and type of residential community (urban, suburban, rural), with race as the independent variable and the outcome, stressor, or prevention behavior as the dependent variables. All analyses were performed using SAS, version 9.4 (SAS Institute).

Of 473 women, 51% (n=241) were White, 13.5% (n=64) API, 12.7% (n=60) Black or African American, 10.1% (n=48) Hispanic or Latinx, 5.7% (n=27) AIAN, and 7% (n=33) multiracial or “other” (Multimedia Appendix 1). The median age was 33 years, with White women significantly older than API, AIAN, and multiracial or other women. Most women had completed some college or graduate school (377/463, 79.7%). Approximately 40% (175/473) of women were unemployed. Approximately 70% (323/471) were in a relationship, and 45% (208/471) had children.

Approximately 46% (211/461) of women lived alone. Of 250 living with others, 55.2% (n=138) and 53.2% (n=133) lived with their partner and family members, respectively. The median number of children in the household was 2. Almost 44% (193/447) had a household income of US $50,000 or more, with White individuals more likely than all racial/ethnic groups, except Latinx. Almost 50% (212/466) of women lived in an urban environment and 17.6% (82/466) in a rural environment.

Approximately 16% (73/470) of women reported COVID-19 symptoms; the highest prevalence was among White (45/240, 18.8%) and multiracial or other (7/33, 21.2%) women, and the lowest prevalence was among API women (4/63, 6.4%; Multimedia Appendix 2). Although 51.4% (227/442) of women knew where to get tested for COVID-19, only 5.8% (27/469) had been tested. White women were more likely to know where to get tested than API, Latinx, and AIAN women (P<.001, P=.01, and P=.01, respectively), and Black women more likely than API and AIAN women (P=.01 and P=.03, respectively). A total of 2% (8/469) of women were diagnosed with COVID-19, of which 50% (4/8) were hospitalized; 22% (18/82) of women were concerned that they were exposed to COVID-19 by the people they know. No significant differences by race/ethnicity were found for either.

A total of 18% (82/469) of women knew someone diagnosed with COVID-19, with White and Black women more likely than API women (P=.02 for both); 67.1% (55/82) identified a diagnosed friend, and 42.7% (35/82) identified diagnosed family. Of the 82 women, 42.7% (n=35) knew someone hospitalized, with Black women more likely than White women (P<.001); family members (16/35, 45.7%) and friends (14/35, 40.0%) were the most frequently reported relationships. Black women were more likely than White women to report a family member hospitalized (P=.01). Of the 82 women, 22% (n=18) knew someone who had died of COVID-19; most were a friend (8/18, 44.4%) or family member (6/18, 33.3%).

Racial/ethnic differences emerged for four COVID-19 stressors (Multimedia Appendix 3). A total of 17% (82/473) of women experienced food insecurity. White women were less likely than API, Black, Latinx, AIAN, and multiracial or other race women to not have enough food (P<.001, P<.001, P=.03, P<.001, and P=.002, respectively); API women were more likely than Latinx women (P=.046) to not have enough food. Approximately 20% (91/473) did not have enough money for rent; White women were less likely to not have enough money for rent than API, Black, Latinx, and AIAN women (P<.001, P<.001, P=.03, and P<.001, respectively). Of 20.9% (99/473) who reported homeschooling children, White women were more likely than API and AIAN women (P=.03 and P=.003, respectively), and Black and Latinx women were more likely than AIAN women (P=.004 and P=.01, respectively). Of 28.1% (133/473) unable to have a doctor or telemedicine appointment, White women were more likely than API and AIAN women (P=.003 and P=.03, respectively). For the remaining stressors, 36.6% (173/473) of women reported loss of income, 15.4% (73/473) were essential employees, 4.7% (22/473) were unable to obtain medication, and 9.9% (47/473) were caregivers for sick family without or with (15/473, 3.2%) COVID-19.

Although no racial/ethnic differences emerged for the Fear of COVID-19 Scale, significant differences were observed for individual items. Higher agreement with the statement “I am very afraid of coronavirus” was found among White women compared to API and multiracial or other women (P=.01 and P=.02, respectively). White women reported more agreement with becoming nervous or anxious when watching media about COVID-19 compared to API women (P=.001). Latinx women agreed with experiencing clammy hands when thinking about COVID-19 more than White and multiracial or other women (P=.02 for both); AIAN women agreed more than White, Black, and multiracial or other women (P=.002, P=.03, and P=.02, respectively). AIAN women agreed more with sleep loss due to worrying about getting COVID-19 compared to White, Black, and multiracial or other women (P=.01, P=.002, and P=.01, respectively); API women agreed more than Black women (P=.02). AIAN women also had more agreement with experiencing their heart racing when thinking about getting COVID-19 compared to Black women (P=.04).

Significant differences in race/ethnicity were observed for four COVID-19 prevention behaviors (Multimedia Appendix 4). Of 95.6% (452/473) of women reporting handwashing with soap, White and API women were more likely than AIAN women (P=.002 and P=.02, respectively). Of 60.0% (284/473) who self-isolated if sick, API women were more likely than White, Black, and multiracial or other women (P=.01, P=.004, and P=.002, respectively). Of 40.2% (190/473) who publicly used gloves, API, Black, and AIAN women were more likely than White women (P<.001, P=.047, and P<.001, respectively); API women were more likely to publicly use gloves than Latinx and multiracial or other women (P<.001 for both); AIAN women were more likely to publicly use gloves than Black, Latinx, and multiracial or other women (P=.02, P=.01, and P<.001, respectively). Of 43.6% (206/473) who did not leave home at all, White women were less likely than API, AIAN, and multiracial or other women (P<.001, P=.001, and P=.04, respectively); API women were more likely to not leave home at all than Black and Latinx women (P=.01 and P=.001, respectively); AIAN women were more likely to not leave home at all than Latinx women (P=.02). Approximately 80% (374/473) of women avoided face touching, 87.1% (412/473) used disinfectants, 88.6% (419/473) stayed home except for essential activities, 82.2% (389/473) covered mouths when coughing, 86.5% (409/473) physically distanced from nonhousehold members, 59.8% (283/473) physically distanced from others if sick, 79.7% (377/473) publicly used a face mask, and 81.8% (387/473) avoided crowds.

To account for potential confounding, COVID-19–related outcomes, stressors, and prevention behaviors that were significantly associated with race or ethnicity in bivariate analyses were identified for inclusion in multivariable binary logistic regressions. Each regression was adjusted for age, education, income, and type of residential community (urban, suburban, rural). Results of these regressions are available in Multimedia Appendix 5 and confirm that there are statistically significant differences in all COVID-19–related outcomes, stressors, and prevention behaviors by race and ethnicity in comparison to the reference group, after accounting for potential confounders.

To our knowledge, this is the first study to describe COVID-19 outcomes, stressors, fear, and prevention behaviors among a sample of US women and to examine racial/ethnic differences. Our data showed that 1 in 6 women had COVID-19 symptoms. Troublingly, although 51.4% (227/442) of women knew where to get tested for COVID-19 and 22.0% (18/82) were concerned about possible exposure to COVID-19, only 5.8% (27/469) had been tested.

White women were the most likely to know where to be tested for COVID-19, despite experiencing a lower burden of COVID-19 among family and friends compared to Black women. This may be due to persistent medical mistrust due to legacies of abuse, mistreatment [22,23], and discrimination as an antecedent to medical mistrust [24]. Social and economic inequalities shape and sustain medical mistrust, particularly among populations who experience staggering health disparities [23]. Further, this may be reflective of an inability to access testing due to structural inequities [10], compounded by employment as essential workers [10,12], limiting available sick leave for testing and support for accessing health care. Given this, it is imperative to engage underserved populations in education and testing through community mobilization, social network, and peer-led engagement approaches.

A total of 18% (82/469) of women knew someone close to them diagnosed with COVID-19—often family or friends—with White and Black women being more likely than API women. Black women were more likely than White women to know someone hospitalized for COVID-19 and report them as family. This demonstrates a high burden of COVID-19 morbidity and mortality among Black families and their social circles; further, this implies a potential burden of adverse mental health associated with loss and trauma.

Racial/ethnic differences were observed for select stressors. White women were less likely than all others to lack food and money for rent, possibly because White women were more likely to have a household income over US $50,000, ensuring the ability to sustain food and housing expenses. White women were also more likely than API and AIAN women to be unable to go to the doctor or have a telemedicine appointment and report homeschooling children. Collectively, plausible explanations are that White women had competing responsibilities due to living with other family members and having a higher median number of household residents, which prevented in-person appointments with their doctor. Additionally, White women may have had privacy concerns with engaging in a phone or video telemedicine visit with their provider due to more family members or household residents being present.

Although significant differences by race/ethnicity were not identified by summed scores for the Fear of COVID-19 Scale, there were significant differences in many of the individual items, showing dimensionality within experiences of fear of COVID-19. White women consistently agreed with statements regarding psychological experiences of fear (being “very afraid” or “nervous or anxious”) relating to COVID-19, while API, Latinx, and AIAN women consistently endorsed agreement with physical manifestations of fear (“hands become clammy,” “cannot sleep,” “heart races”). Black women did not fall into these patterns, endorsing most agreement with being very afraid and experiencing anxiety relating to media about COVID-19. This may reflect differential conceptualization of fear and stress, including antidisclosure socialization and stigma around mental health [25-27] or increased symptomatology associated with adverse mental health resulting from cumulative or toxic stress [11,14].

Overall, we found high prevalence of implementation of CDC-recommended prevention behaviors. Prevalence of staying home except for essential activities was 88.6% (419/473) compared to a rate of 77.3% among US adults surveyed in May 2020 [17]. Likewise, 79.7% (377/473) of women in our study reported using a face mask, slightly higher than 74.1% of US adults surveyed in May [17]. Previous findings demonstrated decreased prevalence in physical distancing behavior from 87% in April to 79.5% in May [16,17]; in our study, 86.5% (409/473) of women reported this behavior. These differences may reflect increased uptake of prevention behaviors among women, compared to a sample of both men and women; this is consistent with findings that female gender is associated with adherence to prevention behaviors [16].

Self-isolation at home if sick was more likely among API women, possibly due to customary traditions of preventing illness. Handwashing with soap was more likely among White and API women than AIAN women. This may be due to White and API women’s increased access to education on preventing COVID-19, primarily disseminated through television and the internet. Of further importance is the possibility of performative practice of prevention behaviors among API women, in response to COVID-19–related xenophobia; individuals of Asian descent have been the targets of hate speech and crime based on the origination of COVID-19 in China [28]. Among AIAN populations, access to the internet is limited [29], and this group is further disadvantaged due to low rates of health literacy [30], limiting knowledge of prevention behaviors. Native patient navigators, similar to community health workers [31], are an avenue for education on COVID-19 prevention in AIAN communities.

API women were more likely than Black and Latinx women to not leave home for any activities, possibly due to API increased partner cohabitation, allowing them to remain indoors and rely on a partner for essential activities. Consistent with existing data, Black and Latinx populations are overrepresented in essential service industries, limiting the ability to shelter at home [10,12]. Because White women had more household residents compared to AIAN women and less partner cohabitation compared to API women, it is plausible that White women were less likely to stay home, due to the need for essential items such as food and medicine.

Supplemental analyses to assess confounding confirmed statistically significant variation between racial/ethnic groups in binary logistic regressions after adjustment for age, education, income, and type of residential community, in comparison to White participants. All adjusted odds ratios were in the expected directions and consistent with bivariate findings. However, the variation in relationship between race/ethnicity and COVID-19–related outcomes, stressors, and prevention behaviors indicate a need for further research to assess the nuanced experiences of women of all racial and ethnic identities, which was beyond the scope of this analysis, to ensure culturally responsive and appropriate public health programming.

The primary limitation of our research was the use of nonprobability sampling methods with recruitment through online social media, resulting in a lack of generalizability of our study findings. Thus, our sample is not representative of adult women in all US states. Our sample was younger to middle-aged, primarily resided in urban environments, and did not consist of racial/ethnic groups and other demographics such as age, education, and income that were proportionate to each US state population’s demographics. Specifically, 17.6% (82/466) of our sample resided in a rural environment, slightly underrepresenting the 19.3% of the US population in rural areas [32]. Although ethnically diverse, our sample underrepresents Latinx women while approximating the US population of Black women. Yet, API and AIAN women are overrepresented in our sample, which may add to the utility of these findings, given the underrepresentation of these groups in research. Finally, only women with access to the internet were able to complete the survey. However, 75% of US adult women use Facebook, with three-quarters accessing Facebook daily [33], indicating that the sampling frame encompassed a significant portion of the US population.

Limitations notwithstanding, our data highlight racial/ethnic differences in COVID-19–related outcomes among adult US women. Our ability to recruit an ethnically diverse sample of women that comprised 41 US states and Washington, DC using online recruitment methods within a short time frame demonstrates the willingness of women to share their experiences on COVID-19. Future studies are needed that use probability sampling methods and longitudinal cohort study designs to examine trends over time in racial/ethnic disparities of COVID-19 outcomes and elucidate underlying factors fueling these disparities among a representative sample of adult US women.

The lower knowledge of testing availability, actualized testing, and prevention behaviors among Black, Latinx, and AIAN women compared to their White and API counterparts will serve to exacerbate the high rates of COVID-19 mortality already demonstrated among these populations. Meanwhile, stressors, physical manifestations of fear, and adverse mental health associated with loss of family and friends to COVID-19 may compound existing toxic stress, culminating in increasingly negative chronic health outcomes. It is vital that public health interventions for COVID-19 focus on these communities, ensuring testing, health care, and support services—including mental health, economic, and social services—are accessible and acceptable, rather than allowing for disparities to persist and worsen.