Racial and Ethnic Differences in Self-Reported COVID-19 Exposure Risks, Concerns, and Behaviors Among Diverse Participants in the Women’s Health Initiative Study

Abstract

Background

Racial and ethnic disparities in coronavirus disease 2019 (COVID-19) risk are well-documented; however, few studies in older adults have examined multiple factors related to COVID-19 exposure, concerns, and behaviors or conducted race- and ethnicity-stratified analyses. The Women’s Health Initiative (WHI) provides a unique opportunity to address those gaps.

Methods

We conducted a secondary analysis of WHI data from a supplemental survey of 48 492 older adults (mean age 84 years). In multivariable-adjusted modified Poisson regression analyses, we examined predisposing factors and COVID-19 exposure risk, concerns, and behaviors. We hypothesized that women from minoritized racial or ethnic groups, compared to non-Hispanic White women, would be more likely to report: exposure to COVID-19, a family or friend dying from COVID-19, difficulty getting routine medical care or deciding to forego care to avoid COVID-19 exposure, and having concerns about the COVID-19 pandemic.

Results

Asian women and non-Hispanic Black/African American women had a higher risk of being somewhat/very concerned about risk of getting COVID-19 compared to non-Hispanic White women and each was significantly more likely than non-Hispanic White women to report forgoing medical care to avoid COVID-19 exposure. However, Asian women were 35% less likely than non-Hispanic White women to report difficulty getting routine medical care since March 2020 (adjusted relative risk 0.65; 95% confidence interval 0.57, 0.75).

Conclusions

We documented COVID-related racial and ethnic disparities in COVID-19 exposure risk, concerns, and care-related behaviors that disfavored minoritized racial and ethnic groups, particularly non-Hispanic Black/African American women.

The coronavirus disease 2019 (COVID-19) pandemic has had a devastating effect across the United States, with the United States reporting more COVID-19 cases and deaths than any other country worldwide (1). Older adults have been affected the most. Individuals 50 years and older account for ~70% of laboratory-confirmed COVID-19 hospitalizations and over 90% of deaths from COVID-19 in the United States (2,3). COVID-19 cases, hospitalizations, and deaths have largely fallen along social fault lines: race, ethnicity, socioeconomic status, and zip code. Non-Hispanic Black/African American and Hispanic/Latinx individuals have been infected at disproportionately higher rates, have higher hospitalization rates, and have 2–3 times higher age-adjusted mortality than non-Hispanic White individuals (4,5). Women also have increased risk of exposure to COVID-19. Women are often responsible for formal and informal caregiving roles that increase their exposure to individuals who are infected with COVID-19 (6). Caregiving responsibilities of women increased during the pandemic as they managed both working and caring for sick family members (7).

COVID-19 was not the great equalizer. Different groups have had a differentiated risk of exposure and differentiated consequences, based on the intersection of multiple marginalized identities, such as gender, race, ethnicity, and social class. For example, women of color make up a majority of the domestic workforce, which is associated with an increased risk of exposure (6). Many reasons have been cited for the disproportionately higher rates of COVID-19 infection and death among non-Hispanic Black/African American individuals, Indigenous individuals, and Hispanic/Latinx individuals. These include higher rates of underlying chronic illnesses, unemployment or employment in public-facing jobs, limited or no health insurance, living in public housing and in multigenerational households, where the risk of acquiring and spreading infection is greater, and living in areas with limited access to COVID-19 testing and resources (8,9). However, these reasons are often cited without reference to the historical context and predisposing policies that laid the foundation for these observations (10). Historically, the United States has created and maintained policies and practices that distribute power and resources unevenly, disproportionately disadvantaging minority groups (11–13).

Race—while being meaningless as a biological marker of difference—is fairly ubiquitous in research data sets and electronic health records and it is a useful marker of exposure to structural racism (14,15). Although there is no validated or universally agreed-upon measure (16), structural racism refers to the totality of ways in which societies foster racial discrimination through mutually reinforcing systems, such as housing, education, employment, criminal justice, and health care. These patterns and practices in turn reinforce discriminatory beliefs, values, and distribution of resources (17).

There are 3 major gaps in the prior literature examining racial and ethnic disparities in COVID-19 exposure and outcomes that this paper seeks to overcome. First, prior studies have generally not examined multiple factors related to COVID-19 exposure, concerns, and behaviors among a large sample of older adults. Second, while some population surveys of older adults early in the COVID-19 pandemic did examine how factors such as having a family member or close friend die of COVID-19 shape risk perceptions and behaviors among older adults (18), they did not examine COVID-19 risk or perception from an intersectional lens of race, ethnicity, sex, and age. Third, prior studies generally have failed to acknowledge race as a social construct and structural racism as the fundamental cause (19) of racial inequities in COVID-19 and other outcomes (20). In other words, structural racism is the reason we see an association between race or ethnicity and worse COVID-19 outcomes in historically marginalized racial and ethnic groups. The current study addresses these gaps by examining multiple factors related to self-reported COVID-19 exposure, concerns, and behaviors among a large sample of older adults from an intersectional lens, guided by a strong conceptual model.

In this manuscript, we present race- and ethnicity-stratified data to examine associations with multiple factors which increased risk of exposure to the severe acute respiratory syndrome- coronavirus 2 (SARS-CoV-2), the virus that causes COVID-19. We also examine how specific COVID-related concerns and behaviors differ in a sample of older postmenopausal women participating in the Women’s Health Initiative (WHI). We specifically acknowledge race as a proxy for structural racism—the root cause of racial and ethnic disparities in COVID-19 and other persistent and pervasive health conditions.

Method

WHI Study

This is a secondary analysis of data from the WHI Extension Study. The original WHI cohort consisted of 161 808 postmenopausal women aged 50–79 years from 1993 to 1998 at 40 clinical centers across the United States (21). Participants were recruited and enrolled onto 1 or more clinical trials (N = 68 132), which tested 3 interventions (menopausal hormone therapy, calcium plus vitamin D supplementation, and/or low-fat dietary pattern) or an observational study (N = 93 676) (22,23). Participants were followed at least annually for vital status and medical outcomes through study closeout (October 2004 to March 2005). After closeout, all participants were invited to reconsent and enroll in subsequent extension studies (2005 and 2010) and those who consented were followed annually. Between June and October 2020 (early in the COVID-19 pandemic), all WHI participants in the extension study who were alive and provided consent to be contacted (n = 64 350) were invited to complete the WHI COVID-19 questionnaire; 49 695 (77.2%) participants did so. Institutional review board approval was obtained by the initial 40 institutions and later by the 4 regional center institutions affiliated with WHI centers and/or the WHI Coordinating Center at the Fred Hutchinson Cancer Research Center.

COVID-19 Questionnaire

The questionnaire included items reflecting risk factors for COVID-19 exposure, such as living arrangement, number of people living in the household, access to visitors, restriction of exit and entry to home, exposure to persons suspected of being COVID-19 infected; death of family or close friends due to COVID-19, COVID-19-related symptoms, COVID-19 testing (frequency, nasal swab, throat swab, saliva test, blood test, test results), hospital stays or treatments for COVID-19, access to medication and health care utilization during the pandemic, degree of concern regarding the pandemic, type and frequency of communication with others outside the home, use of technology to stay in touch with others, alcohol use, smoking, and physical activity. In addition, the questionnaire queried various types of concerns related to COVID-19, including concerns about: social relationships (ability to be with family and friends and risk of family getting COVID infection), safety (risk of getting COVID-19 infection, personal safety, and the health and safety of family and friends), socioeconomic status (having enough money and financial security of self and family), wellness (getting enough exercise and sleep), ability to meet basic needs (getting adequate food, housing, and health care), and the pandemic’s impact on society (the nation and the economy). Table 1 outlines the questions used in this analysis.

Table 1.

WHI COVID-19 Questions and Response Options for Exposures of Interest

Variable	COVID-19 Questions	Question Response Options	Categorization for Analysis and Missing Data
Ever been exposed to someone with COVID-19	To your knowledge, have you EVER been exposed to another person who has been diagnosed with or suspected of having COVID-19 infection?	(1) Yes, someone living with me (2) Yes, someone outside of my household that I interact with face-to-face (3) No, not that I know of	Yes = 1 and 2 No = 3 *1.4% missing data
Family member or friend died from COVID-19	Has anyone in your family or a close friend died from COVID-19?	(1) Yes (2) No	*1.5% missing data
Decided not to go to doctor/hospital to avoid COVID-19 exposure	Have you decided not to go to the doctor or hospital when you normally would have gone to avoid the potential of being exposed to COVID-19?	(1) Yes (2) No	*3.55% missing data
Level of difficulty getting routine medical care since March 2020	In general, how much difficulty have you have getting routine medical care since March 2020?	(1) Some (2) Much (3) Unable or very difficult (4) None	Some/much/very much difficulty = 1, 2, and 3 None = 4 *3.0% missing data
Concern about the COVID-19 pandemic	In general, how concerned are you about the COVID-19 pandemic?	(1) Not at all concerned (2) Somewhat concerned (3) Very concerned	Somewhat/very concerned = 2 and 3 None = 1 *3.4% missing data
COVID-19 pandemic causing concerns about …	Is the COVID-19 pandemic causing you concerns about any of the following? (Mark all that apply)	(1) My risk of getting a COVID-19 infection (2) The risk of family members or friends getting a COVID-19 infection (3) Getting the health care that I need (4) Getting adequate food (5) Getting enough exercise/physical activity (6) Getting the sleep/rest I need (7) Having adequate housing (8) Having enough money to cover my needs (9) My personal safety (10) The health and safety of my family and friends (11) My financial security (12) The financial security of my family (13) My ability to be with friends and family (14) The nation and the economy more generally	Conceptualized as broadly related to social determinants of health (eg, financial security, getting adequate food, getting adequate health care, adequate housing) and concerns about personal safety and safety of family and friends *0% missing data

Notes: COVID-19 = coronavirus disease 2019; WHI = Women’s Health Initiative.

Race and ethnicity

Self-identified race and ethnicity data, the exposure of interest for our analyses, were collected in 2003 based on the 2000 U.S. Census categories: Asian, Hispanic/Latina, non-Hispanic Black/African American, and non-Hispanic White (24). A prior publication details how self-identified race and ethnicity was assessed at baseline in WHI (25). Notably, WHI did not collect race and ethnicity data separately but combined them and then later mapped them to the 2000 U.S. Census categories (25).

Other covariates

Highest educational level and family income were assessed at WHI baseline and neighborhood socioeconomic status (NSES) data were calculated using baseline data. NSES was calculated at the census tract level using an index of 6 variables collected in the 2010 Census: (i) percent of adults older than 25 with less than a high school education; (ii) percent of males who were unemployed; (iii) percent of households with income below the poverty line; (iv) percent of households receiving public assistance; (v) percent of households with children headed by a woman; (vi) and median household income. The index ranges from 0 to 100 across U.S. census tracts, with higher scores indicating more affluent tracts. WHI participants were assigned an NSES index based on their tract of residence.

On annual questionnaires, WHI participants update their medical history, including information regarding body mass index (BMI) and self-reported comorbidities: carotid artery disease, treated diabetes, hypertension requiring pills, history of stroke, history of coronary heart disease (CHD) (myocardial infarction [MI]/coronary artery bypass graft [CABG]/percutaneous transluminal coronary angioplasty [PTCA]), history of (congestive) heart failure (HF), and history of chronic obstructive pulmonary disease (COPD). All data on comorbidities were from WHI baseline and annual questionnaires. BMI was based on the latest measured value for height (2018) and last self-reported weight.

Conceptual Model

Guided by Fundamental Cause Theory (19), we examined how structural racism shapes risk of exposure to COVID-19 (Figure 1). Structural racism operates through long-standing social and economic laws, policies, and practices that have present-day manifestations and determine differential access to resources, risks, and opportunities (26). For example, workforce discrimination leads non-Hispanic Black/African American and Hispanic/Latinx individuals to occupy more public-facing jobs, therefore increasing their risk of exposure (27). Lower socioeconomic status means decreased likelihood of having health care insurance which reduces one’s ability to seek health care if one contracts and experiences complications from COVID-19 (27).

Figure 1.

Conceptual model of associations between structural racism, race, and exposure to coronavirus disease 2019 (COVID-19) virus.

We examined the following predisposing factors that are patterned by structural racism and influence risk of exposure to COVID-19: income, education, geographic region of residence, and comorbidities. The relationship between each of these factors and risk of exposure is socially patterned by race and ethnicity. It is known that minoritized racial or ethnic groups face higher risk of exposure to and severity of illness from SARS-CoV-2 virus due to structural racism. They may have more difficulty getting routine medical care, be more likely to have a family member or friend die of COVID-19, be more concerned about the pandemic, and may be more likely to forego medical care due to fear of COVID-19 exposure.

We hypothesized that, in comparison to non-Hispanic White women, women from minoritized racial or ethnic group in the WHI would:

1. Be at higher risk of exposure based on social and health factors (ie, income, education, and comorbidities) or be more likely to have a family member/friend diagnosed with or die from COVID-19;

2. Have greater difficulty getting routine medical care or decide to forego medical care to avoid exposure to the COVID-19 virus;

3. Report being “somewhat or very concerned” about the COVID-19 pandemic.

We also hypothesized that the racial and ethnic disparities would be most pronounced among non-Hispanic Black/African American women in the sample due to the entrenched history of anti-Black racism in the United States (28), and early data highlighting the disproportionately higher death rates from COVID-19 among non-Hispanic Black/African American adults.

Statistical Analysis

Because the main exposure of interest was race or ethnicity, we created a combined race and ethnicity variable with Hispanic/Latina ethnicity as a separate category; all other single races are presented individually. We conducted a cross-tabulation of race and ethnicity among WHI participants who completed a COVID-19 questionnaire (Table 2). Due to small sample sizes, we dropped the Native Hawaiian/other PI, American Indian/Alaskan native, more than 1 race, and other/not reported racial groups.

Table 2.

Race by Ethnicity Among WHI Participants With a COVID-19 Questionnaire

	Race
	Asian (N = 1 029)		Black/African American (N = 2 791)		White (N = 44 672)		Total (N = 48 492)
	N	%	N	%	N	%	N	%
Ethnicity
Not Hispanic/Latina	1 010	98.2	2 763	99.0	43 647	97.7	47 420	97.8
Hispanic/Latina	19	1.8	28	1.0	1 025	2.3	1 072	2.2

Notes: COVID-19 = coronavirus disease 2019; WHI = Women’s Health Initiative.

We computed descriptive statistics of baseline characteristics by race and ethnicity and examined associations with Chi-square tests of independence for categorical variables or Kruskal–Wallis F tests for continuous variables (Table 3). We examined associations between race and ethnicity (with non-Hispanic White women serving as the reference group) and various COVID-19 risks and concerns (COVID-19 exposure, having a friend/family member die from COVID-19, difficulty getting medical care, and concern about the pandemic) in separate multivariable-adjusted modified Poisson regression models using sandwich estimation for the error variances (29).

Table 3.

Baseline Demographic, Clinical, Psychosocial, and COVID-Related Characteristics by Race/Ethnicity Among COVID-19 Questionnaire Respondents^*

	Overall (N = 48 492)		Asian (N = 1 010)		Hispanic/Latina† (N = 1 072)		Non-Hispanic Black/African American (N = 2 763)		Non-Hispanic White (N = 43 647)
	N	% or Mean (SD)	N	% or Mean (SD)	N	% or Mean (SD)	N	% or Mean (SD)	N	% or Mean (SD)
Age at COVID survey	48 492	83.6 (5.6)	1 010	82.9 (5.7)	1 072	82.3 (5.7)	2 763	82.4 (5.4)	43 647	83.9 (5.6)
<85	28 572	58.9	634	62.8	723	67.4	1 852	67.0	25 363	58.1
≥85	19 920	41.1	376	37.2	349	32.6	911	33.0	18 284	41.9
Education at baseline
High school diploma/GED or less	7 396	15.4	86	8.6	226	21.4	428	15.6	6 656	15.3
School after high school	16 740	34.8	299	29.9	456	43.2	1 015	37.0	14 970	34.5
College degree or higher	24 027	49.9	615	61.5	374	35.4	1 299	47.4	21 739	50.1
Family income at baseline
<$35 000	11 780	25.5	155	16.1	371	36.1	953	36.0	10 301	24.8
$35 000–<$50 000	9 669	20.9	148	15.4	220	21.4	567	21.4	8 734	21.0
$50 000–<$75 000	11 371	24.6	280	29.1	214	20.8	631	23.8	10 246	24.7
≥$75 000	13 358	28.9	380	39.5	223	21.7	496	18.7	12 259	29.5
Neighborhood socioeconomic status quartiles, normalized
≤73.709	10 695	24.6	173	18.9	427	44.6	1 585	68.5	8 510	21.7
73.710–78.430	10 896	25.1	198	21.7	219	22.9	343	14.8	10 136	25.8
78.431–82.222	10 914	25.1	298	32.6	153	16.0	210	9.1	10 253	26.1
>82.222	10 935	25.2	244	26.7	158	16.5	177	7.6	10 356	26.4
Geographic region at COVID survey
Northeast	10 432	21.6	54	5.4	111	10.4	382	13.9	9 885	22.7
South	13 110	27.1	67	6.6	429	40.3	1 458	52.9	11 156	25.6
Midwest	10 910	22.6	34	3.4	56	5.3	594	21.6	10 226	23.5
West	13 923	28.8	854	84.6	468	44.0	321	11.7	12 280	28.2
Comorbidities‡
Carotid artery disease	671	1.4	4	0.4	8	0.7	13	0.5	646	1.5
Treated diabetes	10 602	21.9	265	26.2	314	29.3	1 027	37.2	8 996	20.6
Hypertension requiring pills	34 090	70.3	711	70.4	773	72.1	2 428	87.9	30 178	69.1
History of stroke	2 913	6.0	36	3.6	52	4.9	148	5.4	2 677	6.1
History of CHD (MI/CABG/PTCA)	4 432	9.1	51	5.0	84	7.8	212	7.7	4 085	9.4
History of (congestive) heart failure (HF)§	2 257	4.7	20	2.0	25	2.3	116	4.2	2 096	4.8
History of COPD∥	5 577	11.5	75	7.4	107	10.0	349	12.6	5 620	11.5
Total number of comorbidities
None	10 488	21.6	226	22.4	219	20.4	219	7.9	9 824	22.5
1	21 599	44.5	475	47.0	466	43.5	1 248	45.2	19 410	44.5
2	11 723	24.2	250	24.8	291	27.1	942	34.1	10 240	23.5
≥3	4 682	9.7	59	5.8	96	9.0	354	12.8	4 173	9.6
Body mass index, kg/m2	48 448	26.6 (5.3)	1 010	24.2 (4.3)	1 072	27.4 (5.2)	2 758	29.8 (6.0)	43 608	26.4 (5.2)
Number of people in household
1	20 631	43.9	323	32.8	386	37.7	1 139	42.9	18 783	44.3
2	18 414	39.2	401	40.8	418	40.8	868	32.7	16 727	39.5
3	2 696	5.7	97	9.9	82	8.0	271	10.2	2 246	5.3
4	861	1.8	53	5.4	41	4.0	101	3.8	666	1.6
≥5	1 154	2.5	68	6.9	38	3.7	79	3.0	969	2.3
Not applicable	3 267	6.9	42	4.3	60	5.9	198	7.5	2 967	7.0
Live alone	20 631	43.9	323	32.8	386	37.7	1 139	42.9	18 783	44.3
Ever been exposed to someone with COVID-19
Yes	1 792	3.7	20	2.0	44	4.2	133	4.9	1 595	3.7
No, not that I know of	46 003	96.3	972	98.0	1 011	95.8	2 580	95.1	41 440	96.3
Family member or friend died from COVID-19	2 371	5.0	24	2.4	86	8.1	371	13.7	1 890	4.4
Level of difficulty getting routine medical care since March 2020
None	35 531	75.5	815	83.2	802	76.8	2 047	77.0	31 867	75.3
Some/much/very much	11 499	24.5	165	16.8	242	23.2	611	23.0	10 481	24.7
Decided not to go to doctor/hospital to avoid COVID-19 exposure	11 419	24.4	298	30.7	270	26.2	795	30.1	10 056	23.9
Concern about the COVID-19 pandemic
Not at all concerned	3 150	6.7	59	6.1	88	8.4	124	4.7	2 879	6.8
Somewhat concerned	19 792	42.2	367	37.8	395	37.9	799	30.4	18 231	43.2
Very concerned	23 923	51.0	546	56.2	560	53.7	1 705	64.9	21 112	50.0

Notes: COVID-19 = coronavirus disease 2019; CHD = coronary heart disease; COPD = chronic obstructive pulmonary disease; SD = standard deviation.

^* p Values from a Chi-square tests of independence for categorical variables and from an analysis of variance F tests or Kruskal–Wallis tests for continuous variables were all ≤.0001.

^†Includes participants across all racial groups who identified as Hispanic/Latina.

^‡Includes self-report at baseline as well as adjudicated events during follow-up or self-reported events during follow-up for participants not being adjudicated.

^§Includes self-report of HF at baseline, and adjudicated HF during follow-up for Medical Records Cohort participants or self-reported HF for self-report cohort participants during follow-up.

^∥COPD was only collected as a self-report during Extension Study 2010–2025.

Each model is adjusted for age and geographic region of residence at the time of the COVID-19 survey (minimally adjusted model; Table 4). We considered these variables, a priori, to be confounders in our analyses. We dichotomized age at 85 years because this age group was noted, by the Centers for Disease Control and Prevention, to be at highest risk for COVID complications and death at the beginning of the pandemic (30). Additionally, the association between race or ethnicity and concern about the risk of contracting COVID-19 was examined in a fully adjusted multivariable model (Table 5). The fully adjusted model included age and geographic region of residence at the time of the COVID-19 survey, education level, income, number of comorbidities, having a family member or friend die from COVID-19, and difficulty getting routine medical care during the pandemic. All of these variables were considered to be on the causal pathway (ie, mediators) and reflective of the mechanisms by which structural racism influences risk of exposure to COVID-19. Given the low prevalence of women in our sample who self-reported a history of testing positive for COVID-19 (N = 311, 0.6%) during this early phase in the pandemic, we focused our analyses on more intermediate outcomes that were shaped by one’s risk of exposure to the SARS-CoV-2 virus.

Table 4.

Multivariable Associations Between Race/Ethnicity (exposure) and Various Risks/Concerns Associated With COVID-19

	Non-Hispanic White (reference group; N = 43 647)	Asian (N = 1 010)		Hispanic/Latina (N = 1 072)		Non-Hispanic Black/African American (N = 2 763)
Outcome	N	N	RR (95% CI)*	N	RR (95% CI)*	N	RR (95% CI)*	p Value*
Ever been exposed to someone with COVID-19	42 935	991	0.65 (0.42, 1.01)	1 047	1.16 (0.86, 1.56)	2 705	1.25 (1.05, 1.49)	.01
Family member or friend died from COVID-19	42 930	994	0.72 (0.48, 1.08)	1 048	2.00 (1.63, 2.47)	2 699	3.01 (2.70, 3.35)	<.0001
Difficulty getting routine medical care since March 2020 (some/much/very much vs. none)	42 250	979	0.65 (0.57, 0.75)	1 036	0.94 (0.84, 1.06)	2 650	0.97 (0.90, 1.04)	<.0001
Decided not to go to doctor/hospital to avoid COVID-19 exposure	42 036	970	1.23 (1.11, 1.36)	1 023	1.05 (0.94, 1.16)	2 629	1.25 (1.18, 1.33)	<.0001
Concern about the COVID-19 pandemic (somewhat/very vs. no)	42 123	971	1.12 (1.06, 1.19)	1 035	1.06 (1.00, 1.12)	2 621	1.28 (1.24, 1.32)	<.0001

Notes: COVID-19 = coronavirus disease 2019.

^*Relative risks (RRs), 95% confidence intervals (CIs), and p values are adjusted for age and region of residence (Midwest, Northeast, South, and West) at the time of the COVID-19 survey. Each outcome row represents a separate model. p Values are from Wald Chi-square tests assessing possible differences in the RR by race or ethnicity for each outcome.

Table 5.

Age and Multivariate-Adjusted Associations Between Race/Ethnicity (exposure) and Concern About the Risk of Getting COVID-19

	Non-Hispanic White (reference group; N = 43 647)	Asian (N = 1 010)		Hispanic/Latina (N = 1 072)		Non-Hispanic Black/African American (N = 2 763)
Outcome	N	N	RR (95% CI)	N	RR (95% CI)	N	RR (95% CI)	p Value
Concern about the risk of getting COVID-19 (somewhat/very vs not at all)
Age and region-adjusted*	43 547	1 009	1.20 (1.16, 1.25)	1 064	1.01 (0.91, 1.05)	2 755	1.06 (1.03, 1.09)	<.0001
Multivariate-adjusted†	39 412	913	1.21 (1.17, 1.26)	962	1.04 (0.99, 1.09)	2 466	1.07 (1.04, 1.10)	<.0001

Notes: COVID-19 = coronavirus disease 2019.

^*Relative risks (RRs), 95% confidence intervals (CIs), and p values are adjusted for age and region of residence (Midwest, Northeast, South, and West) at the time of the COVID-19 survey. The p value is from a Wald Chi-square test assessing possible differences in the RR by race or ethnicity.

^†Relative risks (RRs), 95% confidence intervals (CIs), and p value are adjusted for age and region of residence (Midwest, Northeast, South, and West) at the time of the COVID-19 survey, education level, income, number of comorbidities, exposure to someone suspected of or diagnosed with COVID-19, having a family member or friend die from COVID-19, and experiencing difficulty getting routine medical care during the pandemic. The p value is from a Wald Chi-square test assessing possible differences in the RR by race or ethnicity.

Relative risks, 95% confidence intervals (CIs), and p values were calculated for all modified Poisson regression models. All p values are 2-sided with a significance level of .05. We performed 20 tests of significance for differences in relative risk by race or ethnicity; approximately 1 of these would be expected to be significant by chance alone at the .05 level of significance. Additionally, the modeling was performed based on complete case data for each outcome and included covariates; Ns for each model are reported for transparency (Tables 4 and 5). Of note, we excluded women who reported being Native Hawaiian/other PI, American Indian/Alaskan native, more than 1 race, and other/not reported racial groups in Tables 3 and 5. All analyses were performed using SAS for Windows, version 9.4 (SAS Institute Inc., Cary, NC).

Results

The cohort included 48 492 women; only 2.2% (1 072) of the women identified as Hispanic/Latina ethnicity, with the largest proportion (1 025 out of 1 072, 95%) identifying as White race. Among non-Hispanic women, 92% were White, 5.7% were non-Hispanic Black/African American, and 2.1% were Asian (Table 2).

The mean age of the women was 83.6 years old and 44% of these women lived alone. There were significant differences in family income by race and ethnicity. Over 1/3 of non-Hispanic Black/African American and Hispanic/Latina women reported family incomes below $35 000 compared to only 16% Asian and 25% of non-Hispanic White women (Table 3). There were significant differences by race or ethnicity in the total number and types of comorbidities, with non-Hispanic Black/African American being highest among all groups in reporting 3 or more comorbidities.

Overall, only about 3.7% of women self-reported having ever been exposed to someone suspected or diagnosed with COVID. Overall, 5% of women reported that they had a family member or friend die from COVID-19, with this percentage being highest in non-Hispanic Black/African American women (13.7%). Approximately 1 in 4 women overall reported at least some difficulty getting routine medical care since March 2020 and this differed significantly by race and ethnicity (Table 3). While 24.4% of women overall decided not to go to the doctor or hospital to avoid COVID-19 exposure, more Asian (30.7%) and non-Hispanic Black/African American women (30.1%) endorsed this risk mitigation strategy. Furthermore, 93.2% of women reported being somewhat or very concerned about the COVID-19 pandemic and this differed significantly by race and ethnicity.

Multivariable Associations Between Race and Ethnicity and COVID-Related Risks and Concerns

After adjusting for age and geographic region of residence, the relative risk of being exposed to someone with COVID-19 was significantly higher among non-Hispanic Black/African American (adjusted relative risk [aRR] = 1.25, 95% CI: 1.05–1.49) compared to non-Hispanic White women (Table 4). Additionally, non-Hispanic Black/African American, and Hispanic/Latina women were 2–3 times more likely to have a family member or close friend die from COVID-19, compared to non-Hispanic White women. There were significant racial and ethnic differences in the level of difficulty getting routine medical care. Asian women were 35% less likely than non-Hispanic White women to report difficulty getting routine medical care since March 2020 (aRR 0.65; 95% CI 0.57, 0.75). Non-Hispanic Black/African American women (aRR 1.25; 95% CI 1.18, 1.33) and Asian women (aRR 1.23; 95% CI 1.12–1.36) were each significantly more likely than non-Hispanic White women to report forgoing medical care to avoid COVID-19 exposure, after adjusting for age and region of residence. There were also racial and ethnic differences in a variety of COVID-19 pandemic concerns after multivariable adjustment (Table 4). Non-Hispanic Black/African American (aRR 1.28; 95% CI 1.24–1.32) and Asian women (aRR 1.12; 95% CI 1.06–1.19) were both significantly more likely than non-Hispanic White women to be somewhat or very concerned about the risk of getting COVID-19.

After adjusting for age and geographic region of residence, Asian women and non-Hispanic Black/African American women were 20% and 6%, respectively, more likely to be somewhat/very concerned about the risk of getting COVID-19 compared to non-Hispanic White women (Table 5). Furthermore, these results remained true after additionally adjusting for education level, income, number of comorbidities, having a family member or friend die from COVID-19, exposure to someone suspected of or diagnosed with COVID-19, and experiencing difficulty getting medical care during the pandemic.

Discussion

We identified disparities by race and ethnicity in risk of COVID-19 exposure based on social and health factors or having a family member/friend diagnosed with or die from COVID-19, level of concern about COVID-19, and race and ethnic differences in factors which predispose to increased risk of more severe COVID-19-infection and death. Non-Hispanic Black/African American women were at greatest risk of more serious COVID-19 infection based on having a lower socioeconomic status, more comorbidities, and increased exposure to someone suspected of or diagnosed with COVID-19. Non-Hispanic Black/African American women had a higher relative risk of reporting exposure to someone with COVID-19 than non-Hispanic White women, after adjusting for age and geographic region of residence.

The finding that non-Hispanic Black/African American women and Hispanic/Latina were significantly more likely to have a family member or close friend die of COVID-19 compared to non-Hispanic White women is not surprising given that these racial and ethnic groups are known to be higher risk for COVID-19 (31). Although there is limited previous literature on racial and ethnic differences in having a family member or close friend die of COVID-19, a prior study found that non-Hispanic Black/African American women have a higher likelihood of having a family member hospitalized due to COVID-19 (32). Knowing someone who has died of COVID-19 may be indicative of an individual’s COVID-19 exposure risk and may shape an individual’s concerns about COVID-19.

Consistent with other literature (33,34), a larger percentage of non-Hispanic Black/African American women compared with other racial or ethnic groups were most likely to be very concerned about the pandemic. The Health and Retirement Study, a population-based study of racial and ethnic differences in COVID-19 concerns, found that Hispanic/Latinx and non-Hispanic Black/African American adults 50 years and older experience more concerns about the pandemic due to differences in exposure and comorbidities than their non-Hispanic White counterparts (34). However, this study did not examine as broad an array of COVID-related concerns as our study, had a significantly smaller sample size (N = 2 879 older adults), and was less racially and ethnically diverse (eg, did not include Asian participants) than our sample. Another study analyzed the “double jeopardy” of COVID-19 exposure and risk experienced by older non-Hispanic Black/African American people (35). However, this study did not examine other racial and ethnic minoritized groups and did not examine how COVID-19 exposure and risk may vary at the intersection of race, ethnicity, age, and gender. As a result, our study differs from previous research based on our unique sample of racially diverse older women.

Non-Hispanic Black/African American women and Asian women were most likely to forgo medical care to avoid exposure to COVID-19 infection. While we did not formerly test for mediation, it is possible that greater concerns related to the pandemic might explain this finding. Data from prior studies on racial and ethnic disparities in foregoing care are mixed, with some studies showing that non-Hispanic Black/African American individuals are not more likely to report forgoing medical care (36), while others show that non-Hispanic Black/African American and Hispanic/Latinx individuals are most likely to forgo medical care during the pandemic for a variety of reasons, including financial concerns, closure of physicians’ offices, and fear of exposure (36,37).

Forgoing medical care may exacerbate existing chronic health issues and worsen health outcomes long-term for older adults. Minoritized racial and ethnic groups- who already experience more severe chronic illness and at younger ages (38). The implications of this finding should be carefully considered when counseling high-risk groups on COVID-19 risk mitigation strategies. Interestingly, of all the groups, only Asian women were significantly less likely than non-Hispanic White women to report difficulty getting routine care. Taken together with the finding that Asian women had a significantly higher relative risk of deciding not to go to the doctor/hospital to avoid COVID-19 exposure, this could be due to concerns about anti-Asian bias and backlash, which has been documented during the pandemic (39,40). These findings also suggest that decreased health care utilization was as much due to deliberate choices made by certain groups as it was to inadequate access to care.

Like all studies, our study has limitations. First, there is no “gold standard” for accurately measuring exposure to COVID-19 risk. We assessed risk based on risk markers (eg, socioeconomic status, comorbidities, exposure to someone diagnosed with COVID-19). Risk of exposure to COVID-19 may not be accurately measured by factors that explain the disproportionate rates of COVID-19 infection and death in the general U.S. population of women. Moreover, our data are based on participant self-report and therefore subject to recall bias. Second, external validity may be limited because the WHI sample is not representative of the general population of older women in the United States. WHI women tend to be healthier, relatively highly educated, and of higher socioeconomic status. Third, this study was conducted early in the pandemic at a single point in time; however, the risk of exposure changes over time. Repeated assessments over the course of the pandemic would be valuable to examine how these associations change over time.

Despite these limitations, our study has notable strengths, including a large sample size (the largest among published population studies examining racial and ethnic differences in COVID risk and concerns), a racially and ethnically diverse sample, and information on a broad array of COVID-19-specific concerns. In addition, our approach to conceptualization and analyses using race or ethnicity as a primary exposure is aligned with calls from multiple authors and journals who have highlighted the problematic nature of imprecise definitions of race or ethnicity and failure to acknowledge structural racism as a fundamental cause of racial health inequities and who have revised their author instructions accordingly (17,25,41–43,44–46).

Although 1 study contained a conceptually thoughtful explanation of root causes of observed racial and ethnic disparities in COVID-19 exposure, risk, and outcomes (45), we did not find any studies that also provided race-stratified data on COVID-19 exposure risk. Stratification is informative and critical to advancing health equity because it provides quantitative information on whether an exposure has a differential impact on different groups. Our explicit acknowledgment of race as a proxy for structural racism supports Garcia et al.’s framework of structural racism as a fundamental cause of health inequities that result in non-Hispanic Black/African American and Hispanic/Latinx older adults being disproportionately affected by COVID-19 (47).

COVID-19 pandemic and its disproportionate impact on older adults and historically marginalized racial and ethnic groups has sharpened the focus on structural racism and its far-reaching health impact. Structural racism is a form of structural violence because it produces socially unjust conditions that are normalized and reproduced through policies, practices, and laws, predisposing minoritized communities to poorer health outcomes and death (48,49). Structural racism is the driving force behind observed associations between race and ethnicity and COVID-19 and other health outcomes. This paper can serve as a model for other observational studies for how to conduct analyses and report on racial and ethnic health inequities in a thoughtful, theory-driven way that promotes justice (50).

Funding

The WHI program is funded by the National Heart, Lung, and Blood Institute, National Institutes of Health (NIH), U.S. Department of Health and Human Services through contracts HHSN268201600018C, HHSN268201600001C, HHSN268201600002C, HHSN268201600003C, and HHSN268201600004C. T.S.J. is supported by a K23 DK124644-02 from NIH/National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK). K.B. has research funding from National Heart, Lung, and Blood Institute (K01HL142848, R56HL159216, L30HL148881).

Conflict of Interest

None.

Author Contributions

Concept and design: C.W.C., C.P.M. Acquisition, analysis, or interpretation of data: all authors. Drafting of the manuscript: S.J.B., R.P.H., C.W.C. Critical revision of the manuscript for important intellectual content: all authors. Statistical analysis: R.P.H. Obtained funding: J.W.-W.