key: cord-0846862-lt9748vk
authors: Lo, C.-H.; Nguyen, L. H.; Drew, D. A.; Graham, M. S.; Warner, E. T.; Joshi, A. D.; Astley, C. M.; Guo, C.-G.; Ma, W.; Mehta, R. S.; Kwon, S.; Song, M.; Davies, R.; Capdevila, J.; Lee, K. A.; Lochlainn, M. N.; Varsavsky, T.; Sudre, C. H.; Wolf, J.; Cozier, Y. C.; Rosenberg, L.; Wilkens, L. R.; Haiman, C. A.; Marchand, L. L.; Palmer, J. R.; Spector, T. D.; Ourselin, S.; Steves, C. J.; Chan, A. T.; Consortium, COPE
title: Racial and ethnic determinants of Covid-19 risk
date: 2020-06-20
journal: nan
DOI: 10.1101/2020.06.18.20134742
sha: 2ec532d4978d7dc2d9fdef71de82afcd094eb1ad
doc_id: 846862
cord_uid: lt9748vk

Background Racial and ethnic minorities have disproportionately high hospitalization rates and mortality related to the novel coronavirus disease 2019 (Covid-19). There are comparatively scant data on race and ethnicity as determinants of infection risk. Methods We used a smartphone application (beginning March 24, 2020 in the United Kingdom [U.K.] and March 29, 2020 in the United States [U.S.]) to recruit 2,414,601 participants who reported their race/ethnicity through May 25, 2020 and employed logistic regression to determine the adjusted odds ratios (aORs) and 95% confidence intervals (CIs) for a positive Covid-19 test among racial and ethnic groups. Results We documented 8,858 self-reported cases of Covid-19 among 2,259,841 non-Hispanic white; 79 among 9,615 Hispanic; 186 among 18,176 Black; 598 among 63,316 Asian; and 347 among 63,653 other racial minority participants. Compared with non-Hispanic white participants, the risk for a positive Covid-19 test was increased across racial minorities (aORs ranging from 1.24 to 3.51). After adjustment for socioeconomic indices and Covid-19 exposure risk factors, the associations (aOR [95% CI]) were attenuated but remained significant for Hispanic (1.58 [1.24-2.02]) and Black participants (2.56 [1.93-3.39]) in the U.S. and South Asian (1.52 [1.38-1.67]) and Middle Eastern participants (1.56 [1.25-1.95]) in the U.K. A higher risk of Covid-19 and seeking or receiving treatment was also observed for several racial/ethnic minority subgroups. Conclusions Our results demonstrate an increase in Covid-19 risk among racial and ethnic minorities not completely explained by other risk factors for Covid-19, comorbidities, and sociodemographic characteristics. Further research investigating these disparities are needed to inform public health measures.

3 receiving treatment was also observed for several racial/ethnic minority subgroups.

Our results demonstrate an increase in Covid-19 risk among racial and ethnic minorities not completely explained by other risk factors for Covid-19, comorbidities, and sociodemographic characteristics. Further research investigating these disparities are needed to inform public health measures.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 20, 2020. . https://doi.org/10.1101/2020.06.18.20134742 doi: medRxiv preprint

The novel coronavirus disease 2019 (Covid-19) caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) continues to pose a tremendous threat to the global community. As of June 2020, over 6.4 million cases of Covid-19 have been documented worldwide with nearly 381,000 deaths. 1 Prior investigations suggest that Covid-19 disproportionately impacts certain populations, including older individuals, males, 2 and those diagnosed with obesity 3 or other underlying health conditions. 4 However, large-scale studies investigating potential racial or ethnic disparities in infection risk are limited.

Emerging data in the United States (U.S.) and the United Kingdom (U.K.) suggest that racial and ethnic minorities may account for an outsized proportion of Covid-19 hospitalizations and deaths. [5] [6] [7] [8] Although the U.S. Center for Disease Control and Prevention (CDC) recently began mandating the reporting of Covid-19 testing results according to race/ethnicity, 9 data on the risk of testing positive for Covid-19 across a large population are lacking. Most estimates of Covid-19 risk to date are based on reports that are not uniformly collected, rely on a patchwork of information from local authorities, and do not account for other factors that could influence risk, 10 such as comorbidities 3, 11 , ability to practice social distancing, 12,13 income disparity, poorer access to testing/care, and language and cultural barriers. [14] [15] [16] Given the importance of understanding the determinants of health among ethnic minority groups, a comprehensive multinational investigation examining racial and . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 20, 2020. . https://doi.org/10.1101/2020.06.18.20134742 doi: medRxiv preprint ethnic disparities in risk of Covid-19 infection is urgently needed. We conducted a population-scale investigation to examine the risk of reporting a positive SARS-CoV-2 test and presenting for Covid-19 care among racial and ethnic groups in the U.S and the U.K.

We recruited individuals from the general population in the U.S. and the U.K. using the Covid Symptom Study smartphone application ("app") developed by Zoe Global Ltd.

with scientific input from Massachusetts General Hospital and King's College London. 17 The app was launched in the U.K. on March 24 and in the U.S. March 29, 2020 . It offers users a guided interface to report baseline demographic information and comorbidities.

Users are prompted to use the application daily to allow for longitudinal, prospective collection of concomitant symptoms, health care visits, and Covid-19 test results. Study participants were recruited through general media, social media outreach, and direct invitations from the investigators of long-running prospective cohorts. 18 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 20, 2020. . https://doi.org/10.1101/2020.06.18.20134742 doi: medRxiv preprint Information collected through the application has previously been described. 17 Briefly, at enrollment, participants were asked to provide information on demographic factors and suspected risk factors for Covid-19 (Table 1) . On first use and daily, participants were asked if they felt physically normal, and if not, their symptoms, including the presence of fever, persistent cough, fatigue, and loss of smell/taste, among others (Supplemental Table 2 & Supplemental Table 3 ). Participants were asked if they had been tested for Covid-19 and the results (none, negative, pending, or positive). Visits to the hospital for care were documented by participants during daily logs and were recorded if they reported being "in hospital" or "back from hospital". Treatment was recorded if participants indicated receiving any of the following: supplemental oxygen, invasive ventilation, fluids, inhalers, or other treatment.

Individuals were asked to report with which race and/or ethnicity they self-identified.

Questions were based on standard categories from the U.S. National Institutes of Health (NIH) 19 and the U.K. Office for National Statistics. 20 (Supplemental Table 1 ).

Individuals who identified their race or ethnicity as "Other" were provided an option to enter a free-text description. Additional categories created for the U.K. include "East/Southeast Asian" and "Hispanic/Latinx". Individuals who identified as "Mixed Race" or selected more than one race were described as "More than one race" and grouped with "Other". We excluded participants who did not provide information on racial or ethnic identity or selected "prefer not to say". In pooled analyses, categories in the U.S. and the U.K. were harmonized according to NIH categories. 19 "White" and . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 20, 2020. . https://doi.org/10.1101/2020.06.18.20134742 doi: medRxiv preprint "Hispanic/Latinx" were categorized separately and those who were "Non-Hispanic white" were the referent group for all analyses. "Native Hawaiian and Pacific Islanders'', "South Asian", "Chinese", and "East/Southeast Asian" were grouped as "Asian", while "American Indian or Alaskan Natives" and "Middle Eastern" was categorized as "Other" (Supplemental Methods).

Population density was calculated from census data for each Zip Code Tabulation Area 

We employed logistic regression models to examine the odds ratios (ORs) and 95% confidence intervals (CIs) of a positive Covid-19 test and of seeking and receiving treatment. Multivariable models were conditioned upon age, date, and country (if pooled). Additional covariates were selected a priori based on putative risk factors, including sex, body mass index, history of diabetes, heart, lung, or kidney disease, current smoking status, isolation, community interaction with individuals with Covid-19, frontline healthcare worker status, population density, income, and education. Missing . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 20, 2020. . https://doi.org/10.1101/2020.06.18.20134742 doi: medRxiv preprint categorical data were included as a missing indicator. To account for likelihood to receive testing, we performed separate inverse probability weighting (IPW) in the U.S. and the U.K. as a function of race/ethnicity and other factors, such as age, symptom burden, COVID-19 exposure risk factors, and socioeconomic status, followed by inverse probability weighted logistic regression (Supplemental Methods).

In addition to a positive Covid-19 test, we utilized a previously developed symptombased classifier predictive of positive Covid-19 testing. 23 Briefly, using logistic regression and symptoms preceding testing, we found that a weighted score including loss of smell/taste, fatigue, persistent cough, and loss of appetite predicts Covid-19 positivity with high specificity (Supplemental Methods). To further examine the risk of more severe Covid-19, we evaluated the likelihood of Covid-19 and hospital visit or treatment.

We conducted analyses in the U.S. and the U.K. separately but also leveraged harmonized data for pooled results. Two-sided p-values <0.05 were considered statistically significant. All analyses were performed using R 3.6.1 (Vienna, Austria).

Between March 24 and May 25, 2020, 2,414,601 participants (U.S. n = 179,873; U.K. n = 2,234,728) had registered and responded to questions within the app, including 2,259,841 non-Hispanic white, 9,615 Hispanic, 18,176 Black, 63,316 Asian participants, and 63,653 participants of more than one race/other race. Median age was 47 years (interquartile range [IQR] 33-60). Black and Hispanic participants were more likely to be . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 20, 2020. . overweight or obese and have diabetes and kidney disease ( Table 1) . All racial and ethnic minorities reported a lower level of social isolation, were comparatively overrepresented among frontline healthcare workers, and reported a higher likelihood of exposure to individuals with suspected or documented Covid-19. Non-Hispanic white participants more commonly reported living in less densely populated regions, and Black individuals tended to live in locales characterized by lower income and educational attainment. The prevalence of symptoms appeared largely consistent across groups (Supplemental Table 2 & Supplemental Table 3 ). The most common symptoms reported were headache, fatigue, and sore throat.

Among 2,414,601 adults, we documented 10,051 reports of positive Covid-19 testing. In the U.S., all racial minorities had an increased risk of reporting a positive Covid-19 test (age-adjusted ORs ranging from 1.52 to 3.69), with the highest risk among Black participants (Table 2a ). In the U.K., increased risk was observed among Black, South Asian, East/Southeast Asian, Middle Eastern individuals, and those reporting more than one or other race (age-adjusted ORs ranging from 1.23 to 2.85) ( Table 2b) . Results were essentially unchanged with additional adjustment for comorbidities and lifestyle factors. We next performed IPW to account for the likelihood of receiving a test and found modest attenuation of the estimates, with most racial and ethnic minorities remaining significantly associated with a positive test for Covid-19.

Next, we considered the possibility that other social determinants mediated the . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 20, 2020. . association of race and ethnicity with risk of Covid-19 by adjusting for income and education, population density, level of isolation, frontline healthcare worker status, and community exposure to a Covid-19-positive individual (Figure 1) 

To further address the possibility of disparities in access to testing, we used a validated symptom-based model associated with predicted Covid-19 infection in a pooled analysis.

Compared to non-Hispanic white participants, we found a significant increase in risk for 

In both the U.S. and the U.K, compared to non-Hispanic white participants, the aORs (95% CI) for Covid-19 requiring a hospital visit were elevated across racial and ethnic minorities, ranging from 1.32 (1.03-1.68) for more than one/other race to 1.64 (1.20-2.23) for Black race ( CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 20, 2020. .

). Country-specific estimates for combined minority groups compared to non-Hispanic whites were similar to the pooled analysis (Supplemental Table 4 CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 20, 2020. . followed by Asian and mixed ethnic groups in the U.K., even after accounting for demographics, social deprivation, and region 7 . Although these data support a consistently higher risk of worsened Covid-19 outcomes among racial and ethnic minorities, whether the risk of infection also differs by race and ethnicity among the general community warrants further research. Increasing reports from the U.S. CDC, local state health departments, 25, 26 and numerous governmental agencies in Europe suggest a greater risk of Covid-19 among communities of color, 7 but do not generally account for other factors that could influence the risk of infection. 10 Our study complements these early reports, and by accounting for Covid-19 exposure risk factors, sociodemographic information, and comorbidities, our investigation better defines the magnitude of increased risk among ethnic and racial minorities compared to non-Hispanic white participants.

Our results demonstrate that comorbid conditions do not explain the increased likelihood to test positive for Covid-19 among minority populations, especially in the U.S., highlighting the considerable role of structural inequalities in elevating risk. Communities of color may be less able to effectively practice social distancing, 27 given prior literature suggesting they are highly represented among the essential workforce 28 and live in neighborhoods with the higher SARS-Cov-2 infection rates. 29 We were able to show some attenuation of Covid-19 risk when accounting for income, education, population density, measures of social isolation, occupation as a healthcare worker, and exposure to a community member with Covid-19. However, after adjustment, the risk of a positive Covid-19 test remained significant for several racial and ethnic minorities, which is likely . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 20, 2020. . due to additional contributing factors for which we were unable to account, including insurance coverage, access to healthcare, use of public transit, and other essential occupations not specifically queried. Asian and Hispanic populations are also more likely than non-Hispanic whites to live in multigenerational households, 30 and, like Black populations, are more likely to live in densely populated urban areas. 31 Moreover, in the U.S. due to residential segregation, racial and ethnic minorities may live in predominantly minority neighborhoods with higher prevalence of infection, thereby increasing their risk of coming into contact with infected members of the community. 29 Furthermore, Asian and Hispanic populations represent a higher proportion of foreignborn individuals, 30 which poses additional challenges associated with cultural and language barriers, misinformation, immigration-related fear, and anxiety related to accessing care. 32,33 Finally, "weathering" or chronic stress related to structural racism contributes to accelerated aging and chronic diseases that may contribute to COVID-19 risk. 34 The strengths of this study include the use of a smartphone application to rapidly collect prospective data from a large multinational and multiethnic cohort, which offers real-time, actionable Covid-19 risk estimates to inform the public health response to an ongoing pandemic. Second, our study design examined documented, self-reported Covid-19 cases in the general smartphone user population, overcoming limitations related to capturing only more severe cases through administrative hospitalization records or death reports. Third, we examined the risk of predicted Covid-19 according to racial and ethnic groups and found results largely consistent with those of self-reported Covid-19.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 20, 2020. . This approach was not dependent on differences in testing availability which might vary across racial and ethnic groups. Finally, we collected information on and adjusted for a wide range of known or suspected risk factors for Covid-19, which are generally not available in existing registries or population-scale surveillance efforts.

Our study has several limitations. While the use of syndromic surveillance to better understand Covid-19 disparities has great strengths in flexibility, speed and sample size, this methodology is largely dependent upon self-reported data, and therefore susceptible to measurement bias, residual confounding bias, and selection (collider) bias. The probability of app participation, reporting, or access may be differential according to Covid-19 outcomes, minority status and/or covariates. 35 Finally, variables in the app were limited in scope to optimize participation. We acknowledge that the assessed racial and ethnic groups may be oversimplifications, and do not completely characterize the true heterogeneity in how participants experience race and ethnicity.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 20, 2020. . https://doi.org/10.1101/2020.06.18.20134742 doi: medRxiv preprint

In conclusion, within a large population-based sample of individuals in the U.S. and the U.K., we demonstrate a significantly increased risk of Covid-19 and hospital-based evaluation/treatment for Covid-19 among several racial and ethnic minorities compared to non-Hispanic white individuals, which was partially explained by risk factors for exposure and traditional sociodemographic factors. Our results confirm the comparatively outsized burden of Covid-19 on ethnic and racial minorities and the need for further research to understand the basis for these health inequalities.

We express our sincere thanks to all of the participants who entered data into the app, including study volunteers enrolled in cohorts within the Coronavirus Pandemic Epidemiology (COPE) consortium. We thank the staff of Zoe Global Ltd., the

Translational Epidemiology Unit at Massachusetts General Hospital for their tireless work.

Ltd. DAD and ATC previously served as investigators on a clinical trial of diet and lifestyle using a separate mobile application that was supported by Zoe Global Ltd.

Other authors have no conflict of interest to declare.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 20, 2020. 

Data collected in the app are being shared with other health researchers through the . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 20, 2020. . . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 20, 2020. . .

It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 20, 2020. .

It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 20, 2020.

. Figure 1 . Risk of a positive Covid-19 test according to race and ethnicity in the United States and the United Kingdom with additional adjustment for socioeconomic indices. The multivariable association adjusted for comorbidities, as in Table 2 , of race and ethnicity with risk of testing Covid-19 positive in each country is presented (gray). Additional adjustment for isolation (never left home, rarely left home, often left home), frontline healthcare worker (yes/no), community exposure (no, documented, suspected), population density, income, and education in each country, as described in Table 1 , demonstrates attenuation of most associations. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 20, 2020. Adjusted for sex, history of diabetes, heart disease, lung disease, kidney disease, and current smoker status (each yes/no), and body mass index (17-18.4, 18.5-24.9, 25-29.9, and ≥ 30 kg/m 2 ) c Further adjusted for isolation (never left home, rarely left home, often left home), frontline healthcare worker (yes/no), community exposure to Covid-19 (no, documented, suspected), population density, income, and education in each country (each in quintiles).

.

It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 20, 2020. Adjusted for sex, history of diabetes, heart disease, lung disease, kidney disease, and current smoker status (each yes/no), and body mass index (17-18.4, 18.5-24.9, 25-29.9, and ≥ 30 kg/m 2 ). c Further adjusted for isolation (never left home, rarely left home, often left home), frontline healthcare worker (yes/no), community exposure to Covid-19 (no, documented, suspected), population density, income, and education in each country (each in quintiles).

.

It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 20, 2020. 

Johns Hopkins Coronavirus Resource Center

Factors associated with hospital admission and critical illness among 5279 people with coronavirus disease 2019 in New York City: prospective cohort study

High prevalence of obesity in severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) requiring invasive mechanical ventilation

Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study

Racial and ethnic disparities in population level Covid-19 mortality

COVID-19 deaths analyzed by race and ethnicity

COVID-19: review of disparities in risks and outcomes

COVID-19) related deaths by ethnic group, England and Wales -Office for National Statistics

Centers for Disease Control and Prevention. 2020

Does comorbidity increase the risk of patients with COVID-19: evidence from meta-analysis. aging

Not everybody can work from home: Black and Hispanic workers are much less likely to be able to telework

Taking Stock of Essential Workers

Failing another national stress test on health disparities

Implicit racial/ethnic bias among health care professionals and its influence on health care outcomes: a systematic review

Race, socioeconomic status, and health: complexities, ongoing challenges, and research opportunities

For Latinos and Coronavirus, Doctors are Seeing an 'Alarming' Disparity -The New York Times

African Americans are at higher risk of death from coronavirus -The Washington Post

COVID-19 exacerbating inequalities in the US

Socio-spatial factors associated with ethnic inequalities in districts of England and Wales

COVID-19 and African Americans

The United States Census Bureau

Intersection of living in a rural versus urban area and race/ethnicity in explaining access to health care in the United States

Abbreviations: ACE, angiotensin converting enzyme; BMI, body mass index; km, kilometer; kg, kilogram; NSAIDs, non-steroidal anti-inflammatory drugs