COVID-19 infection and sleep health: findings from the nationally representative COVID-19’s Unequal Racial Burden survey

Abstract

Study Objectives:

To estimate overall and racial/ethnic-specific associations between COVID-19 infection status and sleep health.

Methods:

We analyzed online survey data collected from December 2020 to February 2021 among Asian, Black, Latino (English- and Spanish-speaking), and White adults (n = 1,000 each), along with American Indian/Alaska Native, Native Hawaiian/Pacific Islander, and multiracial adults (n = 500). COVID-19 infection (confirmed, probable, suspected), based on self-reported data on symptoms and infected contacts, was classified using World Health Organization definitions. Sleep disturbances were categorized as “yes” (mild/moderate/severe) vs “no” (normal). Weighted analyses were used to generate nationally representative estimates within each racial/ethnic group. Adjusting for sociodemographic and health behaviors, Poisson regression with robust variance estimated prevalence ratios and confidence intervals for sleep disturbances among individuals with vs without a COVID-19 infection in the overall population and by gender, ability to get health care, and race/ethnicity.

Results:

Among 5,359 eligible participants, 24% had a COVID-19 infection. COVID-19 infection was associated with a 32% higher prevalence of sleep disturbances (PR = 1.32 [95% confidence interval: 1.22–1.42]). The higher prevalence of sleep disturbances among women with vs without COVID-19 (PR = 1.32 [1.19–1.45]) was similar to men (PR = 1.34 [1.18–1.53]). COVID-19 infection was associated with an even higher prevalence of sleep disturbances among participants who reported being able vs unable to get needed health care. COVID-sleep associations were higher among American Indian/Alaska Native (PR = 1.64 [1.30–2.09]), Native Hawaiian/Pacific Islander (PR = 1.53 [1.24–1.90]), and English-speaking Latino (PR = 1.49 [1.20–1.86]) adults compared to White adults (PR = 1.14 [0.93–1.41]), although confidence intervals overlapped.

Conclusions:

COVID-19 infections may lead to sleep disturbances among US adults.

Citation:

Alhasan DM, Gaston SA, Strassle PD, Nápoles AM, Jackson CL. COVID-19 infection and sleep health: findings from the nationally representative COVID-19’s Unequal Racial Burden survey. J Clin Sleep Med. 2025;21(6):1053–1063.

BRIEF SUMMARY

Current Knowledge/Study Rationale: COVID-19 infection has been associated with a myriad of symptoms as well as health consequences including sleep problems, likely increasing racial disparities in sleep health. Therefore, we determined the association between COVID-19 infection and sleep health among a racially and ethnically diverse sample in the United States.

Study Impact: COVID-19 infections may lead to sleep disturbances across racial/ethnic groups. Intervention strategies to mitigate harmful effects of pandemics should prioritize individuals with limited access to health care and minoritized racial/ethnic groups who have been disproportionately impacted by COVID-19.

INTRODUCTION

The inequitable disease burden experienced by minoritized racial and ethnic groups due to structural factors (eg, housing segregation,¹ labor market discrimination,² lower insurance coverage³) has been exacerbated by the COVID-19 pandemic.⁴ For instance, minoritized racial and ethnic groups, including Black, American Indian, and Hispanic adults, have been shown to be at higher risk than their White counterparts to contract COVID-19.⁵ Patients with COVID-19 have also been shown to experience a myriad of symptoms as well as health consequences, including sleep problems.⁶ Therefore, the COVID-19 pandemic has likely exacerbated existing sleep disparities,^7–9 which likely increased disparities in subsequent health conditions (eg, cardiovascular disease) that are adversely affected by poor sleep.¹⁰

The multiple sleep disturbances observed throughout the pandemic as a result of the infection or related stressors (eg, economic instability, social confinement) are known as the “coronasomnia” phenomenon.¹¹ In terms of potential biological pathways, a review reported that COVID-19 infection was associated with inflammatory responses, including release of many cytokines, which are important for sleep regulation.¹¹ Patients with COVID-19 have also been shown to have lower absolute lymphocyte count, increased neutrophil-to-lymphocyte ratio, and long duration of hospitalization, which are associated with sleep dysfunction.⁶

Studies report poorer sleep quality throughout each wave of the COVID-19 pandemic,^12,13 including among minoritized racial and ethnic groups compared to White persons,¹⁴ women compared to men,¹⁵ and health care workers compared to the general population.¹⁶ Most studies focused on health care–related stressors during the pandemic and reported associations with poorer sleep health among minoritized racial and ethnic groups.¹⁷ Similarly, reduced access to health care during the pandemic likely increased the risk of negative health outcomes, including sleep disturbances.¹⁸ Health care utilization, especially among individuals with severe illnesses, significantly declined during the pandemic.¹⁹

Nonetheless, few studies have examined the relationship between COVID-19 infection and sleep health among a large, racially/ethnically diverse, and nationally representative sample of men and women in the United States as well as how ability to get health care may modify this relationship. To address research gaps, we aimed to assess the prevalence of sleep disturbances among a racially/ethnically diverse sample as well as to determine associations between COVID-19 infection and sleep disturbances. Specifically, we determined overall as well as gender- and racial/ethnic- specific cross-sectional associations between COVID-19 infection and sleep disturbances. Finally, we assessed whether the ability to get health care modified the association between COVID-19 infection and sleep disturbances among the overall population and by gender and race/ethnicity. We hypothesized the following: (1) the prevalence of sleep disturbances is higher among women compared to men and among minoritized racial/ethnic groups compared to White adults; (2) COVID-19 infection is associated with a higher prevalence of sleep disturbances; (3) the association between COVID-19 infection and sleep disturbances is stronger among women compared to men, among adults unable compared to adults able to get needed health care, and among minoritized racial/ethnic groups compared to White adults; and (4) the magnitude of the association will be higher among racially and ethnically minorized groups of men, women, and the same minoritized individuals whether able or unable to access health care.

METHODS

Data source: COVID-19’s Unequal Racial Burden survey

Participant data for this study was obtained from the COVID-19’s Unequal Racial Burden (CURB) survey, a nationally representative sample of all adults living in the United States. CURB is a cross-sectional, online survey conducted by YouGov, a consumer research firm, from December 8, 2020, to February 17, 2021. The survey was first written in English and translated into Spanish by a certified translator; final versions were produced through team reconciliation by 4 bilingual-bicultural researchers from varied Latino national origin groups. Each study participant provided informed consent. The YouGov research protocol has been approved (FWA00010960) and provides de-identified data, and the National Institutes of Health Institutional Review Board waived approval for data with no identifiable information.

Study population

A detailed description can be found elsewhere.²⁰ Briefly, YouGov uses a proprietary, opt-in survey panel comprising over 1.8 million US residents to conduct nationally representative online surveys. Panel members are recruited using multiple methods to help ensure panel diversity (eg, web advertising, email campaigns, and mail-to-web recruitment) and receive incentives to complete individual surveys through a loyalty program. To obtain nationally representative estimates, YouGov proximity matches eligible panel members using demographic characteristics (matched sample) to a theoretical cohort (target sample). For the CURB study, the 2018 American Community Survey sample was the theoretical cohort. CURB included 1,000 Asian, 1,000 Black/African American (BAA), 1,000 Latino (500 Spanish- and 500 English-speaking, defined as preferred survey language), 1,000 White, 500 American Indian/Alaskan Native (AI/AN), 500 Native Hawaiian/Pacific Islander (NH/PI), and 500 multiracial adults (≥ 18 years old). After surveys were completed, the matched sample was then weighted using propensity scores. Within each racial/ethnic group (and language for Latinos), the matched sample and target sample were combined, and multivariable logistic regression was used to estimate the probability for inclusion in the frame, adjusting for age, gender, years of education, and region. Propensity scores were then grouped into deciles and poststratified on gender, age, education, and region to produce a final weight for each individual in a stratified sample group. This combination of matching and weighting allowed for nationally representative samples within each racial/ethnic group. For this analysis, individuals who identified as transgender or nonbinary were excluded due to small sample size (unweighted n = 128) as well as missing COVID-19 data (unweighted n = 13) with a final analytic unweighted sample of 5,359 participants.

Exposure assessment: COVID-19 infection status

All survey participants were asked, “Since the start of the pandemic, have you had any of the following symptoms that you thought might be related to COVID-19 (check all that apply)?” Options included: (1) fever or chills, (2) cough, (3) shortness of breath or difficulty breathing, (4) fatigue, (5), muscle or body aches, (6) headache, (7) new loss of taste or smell, (8) sore throat, (9) congestion or runny nose, (10) nausea or vomiting, (11) diarrhea, and (12) none of the above. Participants were also asked whether they had been tested for COVID-19 (yes or no); if they reported yes, they were asked about the test result (I tested negative, I tested positive, I don’t know the result). Finally, participants were also asked, “How many people in your household have been infected with COVID-19?” Responses included: 0, 1, 2–4, 5, or more.

Using the self-reported data from these questions, COVID-19 infection status was determined using the World Health Organization guidelines: confirmed/identified, probable, suspected, and none (Table S1 in the supplemental material).²¹ First, all participants who reported having a positive COVID-19 test were classified as a confirmed case. Among individuals without a positive test, those who reported fever/chills and cough or reported 3 or more of the following symptoms—fever/chills, cough, fatigue, headache, muscle or body aches (myalgia), sore throat, congestion or runny nose (coryza), shortness of breath or difficulty breathing (dyspnea), nausea/vomiting, and/or diarrhea—were classified as a suspected case. Participants without a positive COVID-19 test who reported recent loss of taste (ageusia) or smell (anosmia) or who met the criteria for a suspected case and reported having a person in their household having had COVID-19 (contact with probable or confirmed case) were classified as a probable case. For the purposes of this analysis, we dichotomized COVID-19 infection as confirmed, probable, and suspected vs none. In sensitivity analyses, we compared only confirmed COVID-19 infection vs none.

Outcome assessment: sleep disturbances

Self-reported sleep disturbance measures were based on the PROMIS (Patient-Reported Outcomes Management Information System) Short Form v1.0–Sleep Disturbances 4a score.²² PROMIS is a validated measure with a high internal consistency (Cronbach alpha = 0.86). Participants answered 4 questions regarding sleep quality, nonrestorative sleep, sleep problems, and difficulty falling asleep in the past 7 days on a 5-point Likert scale, where higher scores indicated higher sleep disturbance severity. Scores were summed, standardized to a T-score, and categorized based on the PROMIS guidelines: “severe” (T-score > 70), “moderate” (60–70), “mild” (55–60), and “within normal limits” (< 55 [reference]).²² For the purposes of this analysis, we dichotomized sleep as disturbed (severe, moderate, and mild) vs within normal limits.

Potential confounders

Potential sociodemographic confounders considered were age (18–34, 35–49, 50–64, ≥ 65 years), gender (woman; man), race/ethnicity (AI/AN, Asian, BAA, English-speaking Latino, Spanish-speaking Latino, multiracial, NH/PI, White), educational attainment (<high school, high school/General Education Diploma, some college/technical school, ≥ college), region of residence (Northeast, Midwest, South, West), marital/cohabiting status (married/domestic partnership, single/separated/divorced/widowed), nativity status (US-born vs foreign-born), employment status (yes vs no), alcohol consumption during the past month (drinker vs nondrinker), physical activity during the past month (inactive [none, once/mo], underactive [several times/mo to 3 times/wk], active [≥ 4 times/wk]), general physical health (excellent, very good, or good vs fair or poor), and overall mental health (excellent, very good, or good vs fair or poor).

Potential modifiers: gender, inability to access health care, and race/ethnicity

Participants self-identified their gender, inability to get health care, and race/ethnicity. Gender (woman; man) was dichotomized. Inability to get needed health care was also dichotomized as yes vs no based on the following question, “Since the start of the pandemic, was there any time when you did not get medical care that you needed?” Race/ethnicity was based on the question, “Which one of the following would you say best represents your race/ethnicity?” Response options included: Latino/a/x or Hispanic (hereafter Latino), AI/AN, BAA, NH/PI, White, or multiracial.

Statistical analyses

We computed descriptive statistics, and we presented continuous variables as means ± standard deviations and categorical variables as frequencies and percentages. We compared the 2 levels of COVID-19 infection across sociodemographic, health behavior, and clinical characteristics for all participants.

To test associations between COVID-19 infection and sleep disturbances, we used Poisson regression with robust variance to estimate prevalence ratios (PRs) and 95% confidence intervals (CIs) of sleep disturbances (mild/moderate/severe vs within normal limits). The overall model was adjusted for the aforementioned confounders. To test for differences by gender, inability to get health care, and race/ethnicity, interaction terms (eg, COVID-19 infection*inability to get health care) were added to the overall model. We also tested interaction terms (ie, COVID-19 infection*race/ethnicity) in models among men, women, individuals able to get/not seeking care, and individuals unable to get care separately. Analyses were conducted in SAS version 9.4 for Windows (Cary, North Carolina), and a 2-sided P value of .05 was used to determine statistical significance. All analyses were weighted to obtain nationally representative estimates within each racial-ethnic group; counts were rounded for interpretation.

RESULTS

Study population characteristics

Among 5,359 included participants, 23.8% had a COVID-19 infection (Table 1) including 14% identified as confirmed cases, 4.6% as suspected, and 4.9% as probable (data not shown). The mean PROMIS T-score was 51.8 ± 9.6. Approximately 20.0% of participants reported being unable to get needed health care. By race/ethnicity, prevalence of inability to get health care was highest among multiracial (25.3%), NH/PI (21.4%), and Spanish-speaking Latino (21.2%) adults (data not shown). Sleep disturbances were higher among women (40.5%) compared to men (31.6%) and among multiracial (46.3%), AI/AN (44.7%), English-speaking Latino (43.4%), and NH/PI (42.3%) than White (37.3%) adults, whereas sleep disturbances were lower among BAA (33.9%), Spanish-speaking Latino (24.6%), and Asian (28.6%) compared to White adults (Table S2 in the supplemental material).

Table 1.

Study population characteristics, COVID-19’s Unequal Racial Burden survey, December 2020–February 2021, n = 5,359.

Sociodemographic Characteristics	Overall	COVID-19 Infection
		Yes	No
	5,359 (100%)	1,273 (23.8%)	4,086 (76.2%)
Age (years)
18–34	1,927 (36.0%)	502 (39.4%)	1,425 (34.9%)
35–49	1,614 (30.1%)	405 (31.8%)	1,209 (29.6%)
50–64	1,502 (28.0%)	325 (25.5%)	1,177 (28.8%)
≥65	316 (5.9%)	41 (3.2%)	275 (6.7%)
Gender
Men	2,465 (46.0%)	560 (44.0%)	1,905 (46.6%)
Women	2,894 (54.0%)	713 (56.0%)	2,181 (53.4%)
Race/ethnicity
American Indian/Alaska Native	479 (8.9%)	139 (10.9%)	340 (8.3%)
Asian	985 (18.4%)	194 (15.2%)	791 (19.4%)
Black/African American	985 (18.4%)	177 (13.9%)	808 (19.8%)
Latino, English-speaking	491 (9.2%)	127 (10.0%)	364 (8.9%)
Latino, Spanish-speaking	491 (9.2%)	127 (10.0%)	364 (8.9%)
Multiracial	468 (8.7%)	145 (11.4%)	323 (7.9%)
Native Hawaiian/Pacific Islander	478 (8.9%)	113 (8.9%)	365 (8.9%)
White	983 (18.3%)	251 (19.7%)	732 (17.9%)
Nativity status
US-born	4,144 (77.3%)	1,004 (78.9%)	3,140 (76.8%)
Foreign-born	1,215 (22.7%)	269 (21.1%)	946 (23.2%)
Marital status
Married/domestic partnership	2,534 (47.3%)	609 (47.8%)	1,925 (47.1%)
Single/separated/divorced/ widowed	2,825 (52.7%)	664 (52.2%)	2,161 (52.9%)
Employment status
Yes	2,605 (48.6%)	634 (49.8%)	1,971 (48.2%)
No	2,754 (51.4%)	639 (50.2%)	2,115 (51.8%)
Educational attainment
<High school	395 (7.4%)	89 (7.0%)	306 (7.5%)
High school/GED	1,428 (26.7%)	289 (22.7%)	1,139 (27.9%)
Some college/technical school	1,794 (33.5%)	467 (36.7%)	1,327 (32.5%)
≥College	1,742 (32.5%)	428 (33.6%)	1,314 (32.2%)
Region of residence
Northeast	856 (16.0%)	197 (15.5%)	659 (16.1%)
Midwest	835 (15.6%)	211 (16.6%)	624 (15.3%)
West	1,945 (36.3%)	471 (37.0%)	1,474 (36.1%)
South	1,723 (32.1%)	394 (31.0%)	1,329 (32.5%)
Health Behaviors
Alcohol consumption
Drinker	1,574 (29.4%)	339 (26.6%)	1,235 (30.2%)
Nondrinker	3,785 (70.6%)	934 (73.4%)	2,851 (69.8%)
Physical activity (past month)
Inactive	1,668 (31.1%)	426 (33.5%)	1,242 (30.4%)
Underactive	2,459 (45.9%)	597 (46.9%)	1,862 (45.6%)
Active	1,232 (23.0%)	250 (19.6%)	982 (24.0%)

Clinical and Sleep Characteristics	Overall	COVID-19 Infection
		Yes	No
	5,359 (100%)	1,273 (23.8%)	4,086 (76.2%)
General health
Excellent, very good, good	3,995 (74.5%)	821 (64.5%)	3,174 (77.7%)
Fair, poor	1,364 (25.5%)	452 (35.5%)	912 (22.3%)
Mental health
Excellent, very good, good	3,802 (70.9%)	749 (58.8%)	3,053 (74.7%)
Fair, poor	1,557 (29.1%)	524 (41.2%)	1,033 (25.3%)
Inability to access health care status
Unable	1,107 (20.7%)	403 (31.7%)	704 (17.2%)
Able/did not seek care	4,252 (79.3%)	870 (68.3%)	3,382 (82.8%)
Sleep disturbance
PROMIS T-score	51.8 ± 9.6	55.3 ± 9.2	50.8 ± 9.5
Sleep disturbances (PROMIS T-score ≥55)	1,949 (36.4%)	660 (51.9%)	1,289 (31.4%)
Severe (PROMIS T-score >70)	188 (3.5%)	88 (6.9%)	100 (2.4%)
Moderate (PROMIS T-score 60–70)	905 (16.9%)	318 (25.0%)	587 (14.3%)
Mild (PROMIS T-score 55–60)	856 (16.0%)	254 (20.0%)	602 (14.7%)
Within normal limits (PROMIS T-score <55)	3,410 (63.6%)	613 (48.2%)	2,797 (68.5%)

Sleep disturbances categorized based on PROMIS guidelines: “severe” (T-score > 70), “moderate” (60–70), “mild” (55–60), vs “within normal limits” (< 55 [reference]). GED = General Education Diploma, PROMIS = Patient-Reported Outcomes Management Information System.

COVID-19 infection status and sleep health overall, by gender, ability to get health care, and race/ethnicity

Overall, participants with COVID-19 infection vs no infection, had a higher prevalence of sleep disturbances (PR = 1.32 [95% CI: 1.22–1.42]) (Figure 1). These results were similar in our sensitivity analysis comparing confirmed COVID-19 vs no infection (Table S3 in the supplemental material). Minimal differences in the association between any COVID-19 and sleep disturbances were also seen between women (PR = 1.32 [95% CI: 1.19–1.45]) and men (PR = 1.34 [95% CI: 1.18–1.54]) (P_covid*gender = .70) (Figure 1). COVID-19 infection vs no infection was associated with a higher prevalence of sleep disturbances (PR = 1.38 [95% CI: 1.25–1.52]) among participants able to get needed care or who did not seek care than among participants unable to get needed health care (PR = 1.20 [95% CI: 1.05–1.37]); although interaction terms were significant (P_{covid*ability} = .04), the CIs overlapped (Figure 1). Although associations did not significantly differ by race/ethnicity (P_{covid*race/ethnicity} = .39), PRs for associations between any COVID-19 and sleep disturbances were higher among minoritized racial/ethnic groups compared to White adults: AI/AN (PR = 1.64 [95% CI: 1.30–2.09]), NH/PI (PR = 1.53 [95% CI: 1.24–1.90]), English-speaking Latino (PR = 1.49 [95% CI: 1.20–1.86]), Spanish-speaking Latino (PR = 1.41 [95% CI: 0.99–1.99]), BAA (PR = 1.37 [95% CI: 1.15–1.64]), multiracial (PR = 1.29 [95% CI: 1.06–1.58]), Asian (PR = 1.22 [95% CI: 0.97–1.53]), and White (PR = 1.14 [95% CI: 0.93–1.41]) (Figure 2).

Figure 1. Prevalence ratios of sleep disturbances for COVID-19 infection vs no infection among all participants, by gender and ability to seek health care, n = 5,359.

Models are estimating prevalence ratios of sleep disturbances (mild/moderate/severe vs within normal limits). Models are adjusted for age (18–34, 35–49, 50–64, and ≥ 65 years), gender (men, women), race/ethnicity (American Indian/Alaska Native; Asian; Black/African American; Hispanic/Latino, English-speaking; Hispanic/Latino, Spanish-speaking; multiracial; Native Hawaiian/Pacific Islander; White), marital status (married/domestic partnership, single/separated/divorced/widowed), educational attainment (≤ high school, some college, ≥ college), region of residence (Northeast, Midwest, West, South), alcohol consumption (drinker, nondrinker), physical activity (inactive, underactive, active), fair/poor general health (yes, no), and fair/poor mental health (yes, no).

Figure 2. Prevalence ratios of sleep disturbances for COVID-19 infection compared to no infection by race/ethnicity, n = 5,359.

Models are estimating prevalence ratios of sleep disturbances (mild/moderate/severe vs within normal limits). Models are adjusted for age (18–34, 35–49, 50–64, and ≥ 65 years), gender (men, women), marital status (married/domestic partnership, single/separated/divorced/widowed), educational attainment (≤ high school, some college, ≥ college), region of residence (Northeast, Midwest, West, South), alcohol consumption (drinker, nondrinker), physical activity (inactive, underactive, active), fair/poor general health (yes, no), and fair/poor mental health (yes, no).

COVID-19 infection status and sleep health at the intersection of gender and race/ethnicity

Among both women and men, associations between COVID-19 infection and sleep disturbances did not significantly differ by race (P_{covid19*race/ethnicity} = .82; P_{covid19*race/ethnicity} = .18, respectively) and CIs overlapped. Yet, PRs were higher among racially/ethnically minoritized women and men. For instance, associations by race/ethnicity among women were as follows: NH/PI (PR = 1.59 [95% CI: 1.22–2.08]), AI/AN (PR = 1.49 [95% CI: 1.14–1.94]), multiracial (PR = 1.44 [95% CI: 1.08–1.93]), Spanish-speaking Latino (PR = 1.41 [95% CI: 0.87–2.28]), and White (PR = 1.06 [95% CI: 0.82–1.35]) (Figure 3). Results were inestimable among BAA women and among Spanish-speaking Latino and multiracial men.

Figure 3. Prevalence ratios of sleep disturbances for COVID-19 infection vs no infection by race/ethnicity among women and men, n = 5,359.

Models are estimating prevalence ratios of sleep disturbances (mild/moderate/severe vs within normal limits). Models are adjusted for age (18–34, 35–49, 50–64, and ≥65 years), gender (men, women), marital status (married/domestic partnership, single/separated/divorced/widowed), educational attainment (≤ high school, some college, ≥ college), region of residence (Northeast, Midwest, West, South), alcohol consumption (drinker, nondrinker), physical activity (inactive, underactive, active), fair/poor general health (yes, no), and fair/poor mental health (yes, no).

COVID-19 infection status and sleep health by inability to get health care across race/ethnicity

Among individuals able to get health care/did not seek care, associations between any COVID-19 infection and sleep disturbances did not significantly differ by race/ethnicity (P_{covid*race/ethnicity} = .58) (Figure 4). Among participants unable to get health care, associations between COVID-19 infection and sleep disturbances varied by race/ethnicity (P_{covid*race/ethnicity} = .04). For instance, COVID-19 infection was associated with a higher prevalence of sleep disturbances among BAA (PR = 1.50 [95% CI: 1.13–1.98]) adults, which differed from the marginal association with a lower prevalence among multiracial adults (PR = 0.77 [95% CI: 0.57–1.05]). There were marginal associations among NH/PI (PR = 1.63 [95% CI: 1.00–2.66]) and Asian (PR = 1.47 [95% CI: 0.95–2.26]) adults and no association among White (PR = 0.93 [95% CI: 0.71–1.20]) adults. Associations for English- or Spanish-speaking Latino adults with inability to get health care were inestimable.

Figure 4. Prevalence ratios of sleep disturbances for COVID-19 infection vs no infection by race/ethnicity and ability to get health care, n = 5,359.

Models are estimating prevalence ratios of sleep disturbances (mild/moderate/severe vs within normal limits). Models are adjusted for age (18–34, 35–49, 50–64, and ≥ 65 years), gender (men, women), marital status (married/domestic partnership, single/separated/divorced/widowed), educational attainment (≤ high school, some college, ≥ college), region of residence (Northeast, Midwest, West, South), alcohol consumption (drinker, nondrinker), physical activity (inactive, underactive, active), fair/poor general health (yes, no), and fair/poor mental health (yes, no).

DISCUSSION

Among this nationally representative sample of US major racial/ethnic adults, COVID-19 infection was associated with more sleep disturbances. The prevalence of COVID-19 infection was substantial among our sample (24%). We also found that the prevalence of sleep disturbances was higher among women compared to men and among certain minoritized racial/ethnic groups (ie, multiracial, AI/AN, English-speaking Latinos, and NH/PI) than White adults, whereas other groups (ie, BAA, Spanish-speaking Latino, and Asian) had a lower prevalence than White adults. Consistent with our hypothesis, overall, COVID-19 infection was associated with sleep disturbances in regression analyses. Our sensitivity analyses restricting to confirmed/identified vs no COVID-19 infection also corroborated the hypotheses. However, we did not find differences by gender, and there was no evidence of significant effect modification by race/ethnicity among the overall population. Within gender groups, COVID-19 infection vs no infection had a higher yet nonstatistically significant association with sleep disturbances among certain minoritized racial/ethnic groups compared to their White counterparts. Finally, among individuals unable to get health care, we observed stronger associations among minoritized racial/ethnic groups (except multiracial adults) compared to White adults, which was consistent with our hypothesis.

Previous studies also report poorer sleep health among adults and minoritized racial/ethnic groups. A systematic review of 16 studies reported a high prevalence of sleep disturbances among patients with COVID-19 and discharged COVID-19 survivors.²³ A meta-analysis of 250 studies also reported higher levels of sleep disturbances during the pandemic in general population samples, health care workers, patients with COVID-19, children, and adolescents.¹³ Investigating racial/ethnic-related disparities in sleep health during the pandemic, a study reported more insomnia symptoms among BAA compared to White adults,²⁴ and a second study reported that BAA young adults had the shortest sleep duration and worst sleep quality compared to White adults.²⁵ Another study reported that Asian Americans who witnessed or heard about other individuals of their racial group being the target of racism during the pandemic was associated with poorer sleep quality and shorter sleep duration.²⁶ Among college students in the southeastern US, another study reported more sleep problems among women than men and among BAA men than White men.¹⁵ Although the higher prevalence of sleep disturbances among minoritized racial/ethnic groups than White adults observed in our study may be due to sleep differences before the pandemic, previous research suggests that sleep disturbances were exacerbated during the pandemic.²⁴

Although there is limited research examining associations between COVID-19 infection and sleep health, previous studies report associations between more COVID-19-related stress and poorer sleep health. Among young adults, a prior study showed that COVID-19-related worries were associated with more sleep problems, measured via the Medical Outcomes Study Sleep Scale.²⁷ Another study also using CURB data found that financial hardship during the pandemic was associated with more sleep disturbances, particularly among BAA adults.²⁸ Our results demonstrate associations between COVID-19 infection and more sleep disturbances among diverse racial/ethnic groups, such as AI/AN, NH/PI, and Latino (regardless of language spoken) groups. COVID-19 infection may lead to sleep disturbances through direct and indirect mechanisms. For example, COVID-19 infection is associated with inflammatory responses, including the release of various cytokines as well as chemokines (eg, interleukins), which can induce changes in circadian rhythm and thus disrupt sleep.^6,11 COVID-19 infection may also lead to sleep disturbances through psychological distress by, for example, financial strain, anxiety, or isolation.²⁹ This may be exacerbated among minoritized racial/ethnic groups who are more likely to receive lower wages, be employed in jobs where remote work is not possible, and live in multigenerational households where isolation to prevent infection may be more difficult.²⁹

We reported nonstatistically significant heterogeneity in associations between COVID-19 infection and sleep disturbances among racially/ethnically minoritized women compared to White women. Specifically, we observed higher PRs among NH/PI, AI/AN, multiracial, and Spanish-speaking Latino women compared to White women, although CIs overlapped. Our findings are similar to previous research that reported minoritized racial/ethnic groups of women—Asian, NH/PI, Latino, and AI/AN—not being able to sleep during the pandemic.³⁰ Another prior study among Latino transgender women found that sleep problems were associated with COVID-19-related stressors (eg, job loss).³¹ These findings contribute to the literature by demonstrating the intensified experience of race/ethnicity (racism) and gender (sexism) on health.

Ability to get/did not seek needed health care modified the association between COVID-19 infection and sleep disturbances in our sample. These results were expected, given how structural barriers impede health care access (as well as utilization) among minoritized racial/ethnic groups.³² For example, a recent study found that hospitals were assigned a lower dollar value (based on patient care revenues and profits) if they had more BAA patients.³³ Moreover, minoritized racial/ethnic groups are overrepresented in low-wage jobs that lack health benefits and employer-provided health insurance.³⁴ In fact, a systematic review including 81 studies from 20 countries found a 37% average reduction in health care utilization during the pandemic, although race/ethnicity was not evaluated.¹⁹ A previous study using the CHASING COVID Cohort found that among adults with higher COVID-19 susceptibility, minoritized racial/ethnic groups—Hispanic, BAA, Asians (which also included AI/AN)—compared to White adults, had a higher prevalence of hospitalization, although there was no association between health care access and hospitalization.³⁵ Another study in Ireland found that older adults with 2 or more chronic conditions compared to none were more likely to have delayed health care utilization.³⁶ Given prior documentation of systematic maltreatment of minoritized racial/ethnic groups resulting in substandard care,³² future research is warranted. Residential segregation, along with other institutional factors, shape health care access, use, and quality where often lower-quality facilities with fewer providers are available in predominantly BAA neighborhoods.³⁷

Intervention strategies to mitigate harmful effects of pandemics should prioritize minoritized racial/ethnic groups who have been disproportionately impacted by COVID-19. Because pathogenic effects of exposure to adverse living and work conditions increase risk for COVID-19 and exacerbate other health conditions,³⁸ interventions must go beyond the individual level and address structural inequities. For example, there is a need to address indoor pollutants and improve housing.³⁹ BAA, Latino, and AI/AN groups are more likely to live in crowded conditions and in multigenerational households, which makes quarantining to reduce COVID-19 risk difficult. Similarly, minoritized groups, particularly BAA and Latino women, are more likely to work in retail (eg, restaurant service), where COVID-19 exposure is higher due to overcrowding, poor ventilation, and limited mask-wearing as well as being deemed an “essential” worker. For instance, BAA and Latino adults comprise 25% of service industry workers compared to 16% for White workers.⁴⁰ Improving health care access may include increasing health insurance coverage among BAA and other minoritized racial/ethnic groups, who tend to have the lowest rates of insured status.³⁸ This may also increase access to testing as often testing is obtained through a primary care clinician. Likewise, improving quality of health care is needed, which may be another avenue to improve access to testing. Studies show that minoritized racial/ethnic groups have longer waiting times and are less likely to be appropriately diagnosed and treated for pain.⁴¹

Study limitations include self-reported data on COVID-19 infection and sleep disturbances, which may introduce measurement error. Subjective sleep disturbances may be underestimated, particularly among minoritized racial/ethnic groups^42–44; however, we used a validated measure (ie, PROMIS).²² Further, the measurement of COVID-19 infection, although based on the World Health Organization guidelines, used self-reported symptoms and COVID-19 test results, and does not capture individuals who are asymptomatic. Nonetheless, our calculated prevalence of COVID-19 infection corroborates with national estimates from that time, except among BAA adults where we calculated a lower prevalence.⁴⁵ This may be attributed to sampling bias (eg, online survey, limited internet availability), survival bias (eg, BAA had higher COVID-19-related deaths and thus may not have been captured in the survey), or structural barriers (eg, limited testing centers available in predominantly poor, BAA neighborhoods).⁴⁶ It is important to collect comprehensive data on COVID-19 deaths by race/ethnicity and other sociodemographic characteristics (eg, age, nativity, gender), as these factors influence health outcomes.³⁴ Similarly, because the survey was only administered online and in English and Spanish, this may also contribute to sampling bias, particularly among Asians where 12.8% in our study had limited English proficiency. Moreover, the small percentage of individuals unable to get needed health care may limit our power to detect effect modification. Finally, the cross-sectional design limits our ability to infer causality.

Study strengths include the use of a racially/ethnically diverse sample allowing us to examine the relationship across many racial/ethnic groups, such as AI/AN, NH/PI, and multiracial, who are understudied. The CURB survey is also generalizable to the US population representing all major racial/ethnic groups in the United States, even though the CURB distribution does not reflect the actual distribution of racial/ethnic groups in the US census. Also, the survey was conducted during a major peak in the COVID-19 pandemic (ie, December 2020–February 2021), which is important to capture longer-term impacts of the pandemic on sleep and identify individuals who are able and unable to recover from sleep disturbances associated with the pandemic.⁴⁷

In conclusion, COVID-19 infection was associated with sleep disturbances across racial/ethnic groups in the United States and demonstrates the need for “upstream” interventions to address ongoing health disparities. Interventions aimed at equitably preventing infection as well as improving quality and increasing access to care are a first step to addressing structural health inequities. Given the high burden of COVID-19 infection and more sleep disturbances among certain minoritized racial/ethnic groups, policies related to dismantling structural barriers may help mitigate the disproportionately higher disease burden among these populations.

DISCLOSURE STATEMENT

All authors have seen and approved the manuscript. This work was funded by the Intramural Program at the National Institutes of Health (NIH), National Institute of Environmental Health Sciences (Z1AES103325), and by the Division of Intramural Research at the NIH, National Institute on Minority Health and Health Disparities (C.L. Jackson). The authors report no conflicts of interest.

ABBREVIATIONS

AI/AN

American Indian/Alaska Native

BAA

Black/African American (non-Hispanic)

CI

confidence interval

CURB

COVID-19’s Unequal Racial Burden

NH/PI

Native Hawaiian/Pacific Islander

PR

prevalence ratio