Racial disparities in COVID-19 deaths in Georgia

Abstract

The 2019 novel coronavirus disease (COVID-19) has brought to the forefront racial disparities in health outcomes across the US, but there is limited formal analysis into factors associated with these disparities. In-depth examination of COVID-19 disparities has been challenging due to inconsistent case definition, isolation procedures, and incomplete racial and medical information. As of June 2020, over 14,000 (25%) confirmed COVID-19 cases in Georgia did not have racial information. However, nearly all COVID-19 deaths had racial and ethnic information for analysis. Using county-level information from the Georgia Department of Public Health and the national County Health Rankings & Roadmaps, we found that Black Americans represented 31.5% of all Georgia residents but 46% of COVID-19 deaths. In the metropolitan Atlanta area, this over-representation was most pronounced in Fulton County which houses the City of Atlanta. The opposite pattern – worse disparity in counties surrounding the central city-bearing county – was instead observed in Albany, Columbus, and Macon, with no significant disparity difference in counties surrounding Savannah. Principal component analysis of health-related outcomes and social determinants of health from these 46 counties identified 17 themes, with greater racial disparities in COVID-19 deaths associated with worse air pollution, more rural communities, and paradoxically greater adherence to guidelines for screening mammography. We conclude that factors associated with the virus responsible for COVID-19 and healthcare disproportionately impact Black Americans.

Introduction

The number of deaths related to the 2019 novel coronavirus disease (COVID-19) has risen quickly in the US from 12 during February to 43,300 during May 2020. COVID-19 is caused by the beta coronavirus SARS-CoV-2 which is structurally similar to the original coronavirus related to the 2003 outbreak of severe acute respiratory syndrome (SARS). Fewer than 10 confirmed 2003 SARS cases were reported in the US, with most North American cases occurring in Toronto involving significant hospital-based transmission (Booth et al., 2003; Centers for Disease & Prevention, 2003a, 2003b; Dwosh, Hong, Austgarden, Herman, & Schabas, 2003). In the current pandemic, COVID-19 has affected more than 1% of the population in 27 states. Racial disparities in COVID-19-related hospitalization were highlighted early in Georgia focused on mostly hospitals from Atlanta(Gold et al., 2020). However, racial information is missing from 24% of confirmed cases in Georgia which makes “Unknown” the most common racial group after non-Hispanic whites (White, 34%) and Black/African Americans (Black, 30%; (“Georgia Department of Public Health Daily Status Report,” 2020). This “missingness” is not unique to Georgia, as the 27 most affected states have a median of 26% missing racial information(“The COVID Racial Data Tracker,”). Hence, while Black Americans have been found to bear a disproportionate proportion of some COVID-19 burden in selective locales, incomplete information on race and ethnicity in many public health reports likely obscure the full scale of these disparities.

In contrast to confirmed cases, most states are more likely to report racial information in deaths due to COVID-19 (median of 4% missing data; (Millett et al., 2020). Because Georgia accounts for the top four counties across the US with the highest COVID-19 deaths per capita(“The COVID Racial Data Tracker,”), it is critical to determine if Black Georgians – at the county level – bear a disproportionate burden of COVID-19-related mortality. Early reports identified two COVID-19 epicenters in Georgia connected by a funeral attended by residents from the greater metropolitan Atlanta area and the City of Albany in southcentral Georgia,(Barry, 2020) and they continue to have the greatest per-capita COVID-19 cases within the state. In the current work, we combined COVID-19 death data for key counties from the Georgia Department of Health, and interrogated their relationships to indicators of county residents’ physical well-being, access to healthcare, environmental quality, and racial segregation.

Methods

Data sources

Publically available, de-identified data from the Georgia Department of Health, US Centers for Disease Control and Prevention, and County Health Rankings and Roadmaps (www.countyhealthrankings.org) were used in this study. Mortality data related to COVID-19 were downloaded on 6/12/2020, with information on race, age, and county of residence. County Health Rankings (CHR) focus on community-level and environmental health-related outcomes(Hendryx, Ahern, & Zullig, 2013; Peppard, Kindig, Dranger, Jovaag, & Remington, 2008; Wagner et al., 2012). The CHR was first refined in the state of Wisconsin and subsequently expanded to other states starting in 2008 (Remington, 2015). Measures in CHR focus on those related to population health, potentially modifiable by individual practices or public policies, assessed by reliable tools, regularly updated, and supported by content experts. These health indicators include 1) demographic factors: proportion of residents according to race, sex, age groups (<18 and >65), rural, proficient in English, high school completion, some college education, unemployment; 2) socioeconomic factors: median household income (partly by race for some counties), % home ownership, % with severe housing cost burden, residential segregation index between Blacks and Whites or between non-Whites and Whites, % children in poverty (partly by race), children in single-parent household, income inequity between 80th and 20th percentile, % without medical insurance; 3) health outcomes: life expectancy (by race), premature death measured by year potential life lost (partly by race), prevalence of adults with poor-to-fair health, average days of poor physical and mental health per year, prevalence of adults smoking, diabetes, obesity prevalence, HIV prevalence, chlamydia prevalence, prevalence of low birth weight, and rates of infant and child mortality, teen births (by race), injury deaths, drug-overdose deaths (excluded due to missing data), motor vehicle deaths, suicide, firearm fatalities, homicides; 4) access to healthy behaviors: food environment index, and proportion with food insecurity, access to healthy food, physically inactivity, access to exercise opportunities, heavy alcohol consumption, flu vaccination (partly by race), and insufficient sleep; 5) structural and environmental factors: number of primary care physicians, dentists, and mental health providers; rates of preventable hospitalization (partly by race), mammography screening (partly by race); 5) environmental factors: air pollution measured by particular matter 2.5 [PM2.5], water quality measured by violations, and proportion of residents in overcrowded housing. Georgia-specific ranking data was downloaded on 6/12/2020.

In addition to these measures, a racial disparity score was calculated for each variable with available breakdown according to race by dividing the in-county value for Black residents by the in-county value for White residents. For example, Black-to-White preventable hospitalization score is calculated by dividing the preventable hospitalization rate for Black residents by the rate for White residents in that county. For disparity in COVID-19 deaths within each county, proportion of COVID-19 deaths reported by DPH as Black was divided by proportion of residents reported by CHR as Black, and the difference from 1 (unity) was used as a surrogate measure of racial disparity.

Statistical Analysis

All statistical analysis was performed in IBM SPSS 26 (Armonk, NY). To characterize COVID-19-related deaths in Fulton and surrounding counties, age, sex, and race information for deaths reported by the Georgia DPH were compared between counties using Chi-squared or Fisher’s exact test for categorical variables and Student’s T-test for continuous variables. For each of the other city-bearing counties, surrounding counties had fewer COVID-related deaths and were analyzed as a group (e.g., between Dougherty and its surrounding counties for Albany).

To identify independent variables predictive of racial disparities in COVID-19-related deaths, we first used principal component analysis (PCA) to identified themes in social determinants of health and health outcomes from CHR(Conroy et al., 2018; Eibner & Sturm, 2006; Morris & Carstairs, 1991). PCA is a commonly used technique to distill large number of variables – many of which are likely inter-related based on theoretical ground, measurement tools, biological or social forces capable of influencing multiple outcomes – down to manageable number of independent dimensions or principal components (PC). Variables associated with each PC were examined for the common theme within the PC, as the nominal measure (e.g., number of residents per mental health provider in a county) may actually reflect another trait (e.g., rural community) beyond its face validity. For simplicity, the strongest contributor identified via VariMax rotation (rather than the calculated PC value) was used in subsequent regression models. The 17 factors identified were first compared between five central counties and all the peripheral counties using Student’s T-tests. They were then analyzed (also using Student’s T-tests) between the counties in the greater Atlanta area and other areas. Benjamin-Hochberg procedure was used for each to adjust for multiple comparisons with false discovery rate (FDR) < 0.10.

Factors different between groups were then entered into a logistic regression analysis to determine their relationships with greater racial disparities in COVID-19-related deaths. In the first model, greater Atlanta and central-vs-peripheral county status were included to identify additional variables influencing the likelihood of disparity. In the second model, greater Atlanta and central-vs-peripheral county status were replaced with health-related and social determinants of health factors. In addition, linear regression analysis was used to identify factors associated with a continuous measure of racial disparity for each county (proportion COVID-19 deaths reported as Black/proportion of residents reported as Black – 1). There were four counties (Chattahoochee near Macon, Crawford and Jones near Macon, Evans near Savannah) without any COVID-19 deaths. These four counties were included in the logistic regression analysis as a peripheral county, but were excluded from the linear regression analysis because no county-specific death rate could be calculated.

For logistic and linear regression models, Z-transformed values for each variable was used based on Georgia-specific mean and standard deviations: 61.91±29.0% for %rural, 34.43± 5.87% for % with obesity, 12.12±1.61% for % with frequent physical distress, 10.74±0.62 for daily average PM2.5, 1.186 ±0.483 for Black:White preventable hospitalization rate, 68.18±20.72 for child mortality rate, 77.90±17.37 for injury death rate, 40.0±5.14 for percent women (65–74) with screening mammography, 5325±1333 for preventable hospitalization stays per 100,000 Medicare enrollees, and 1.248±0.351 years of potential life lost for Black:White premature death.

Results

Across the state of Georgia, 31.5%, 9.8%, and 4.3% of residents self-reported as Black (non-Hispanic), Hispanic, and Asian (non-Hispanic) in 2020. As of mid-June, the same groups accounted for 46%, 4.8%, and 1.5% of all recorded COVID-19 deaths in Georgia (p<0.001 compared to state population). Based on these numbers, Black Georgians were 75% (95% CI 62 – 89%) more likely to have died from COVID-19 than White Georgians. This is quite different from the comparable rates of influenza-related deaths between Black and White Americans (Hutchins, Fiscella, Levine, Ompad, & McDonald, 2009).

As of mid-June, greater Atlanta and City of Albany continued to have the highest number of COVID-19 cases and deaths (Fig 1). All city-bearing counties had greater deaths than surrounding counties, but only Fulton County had much greater representations of Black Georgians in its COVID-19 deaths than expected based on proportion of Black in-county residents. This disparity was not observed in any of the ten counties surrounding Fulton, but was conversely observed in the counties surrounding Albany, Columbus, and Macon (but not Savannah).

Figure 1.

Patterns of racial disparities in COVID-19-related deaths for communities surrounding five major Georgia cities A: Central city-bearing counties (Atlanta in Fulton County; Albany in Dougherty County; Columbus in Muskogee County; Macon in Bibb County; Savannah in Chatham County) and their surrounding counties (gray) are shown. B: Representation of Blacks in their county population and COVID-19-related deaths are shown relative to unity line (dotted). Size of circles correspond to number of total COVID-19-related deaths, and color coding is shared between A and B.

To explore whether health-related measures and social determinants of health contributed to not only the disparity between Fulton and other city-bearing central counties but also between the central and surrounding counties, we first examined these factors from the CHR. Because of the high number of variables available from CHR and the expected likelihood that some or many will correlate with others, we performed PCA to identify themes measured by CHR beyond the nominal measures (see Methods). In the counties examined in this study, 17 PC (with their representative variables and loading score, Figure 2) were identified: poverty (% with frequent physical distress, 0.940) ; rural community (no. of residents per mental health provider, 0.743); racial inequity in life expectancy (Black:White ratio of years of potential life lost); injury death (injury death rate, 0.815); access to care (% uninsured, 0.778); community safety (homicide rate, 0.740); motor vehicle mortality (motor vehicle mortality for Black Georgians, 0.852); disposable income (number of dentists in county, 0.722); air pollution (daily average PM2.5, 0.956); obesity-related illnesses (% with obesity, 0.774); family and social support (social associations, 0.671); healthcare inequity (Black:White ratio of preventable hospitalization rate, 0.884); child poverty inequity (Black:White % children in poverty, 0.691); societal disconnection (% not proficient in English, 0.765); quality of care (preventable hospitalization rate, 0.719); preventive care (% women 65–74 years with screening mammography); regulation enforcement (water quality violation, 0.602);

Figure 2.

Principal component analysis of factors considered in the current study. Each column represents a principal component used for further analysis. Colors represent strengths of associations between each factor (row) and principal component, with red indicating positive association and blue indicating negative association.

As central and peripheral counties showed different trends of racial disparities in COVID-19 deaths across the five major areas, we first compared the 17 newly identified variables between central and peripheral counties (Figure 3). This only identified number of residents per mental health provider to differ between the two groups (Table 2A). Because counties in the greater Atlanta area also showed greater disparity centrally, we then compared these counties (n=10) against the remaining Georgia counties included (n=32). Compared to other counties, greater Atlanta counties had fewer or lower residents in rural areas, preventive care, injury death rates; and more or greater poverty, pollution, quality of care, greater community safety (Table 2B, FDR<0.10 for all).

Figure 3.

Comparison of key variables between counties surrounding five Georgia cities. Central city-bearing counties are colored red (greater racial disparity in COVID-19-related deaths), blue (less disparity), or empty (similar disparity) according to comparison with surrounding counties.

Table 2.

Factor differences between ATL counties and non-ATL counties.

A: Differences between counties in the metropolitan Atlanta area and other counties included in analysis. B: Differences between counties surrounding Fulton and other city-bearing counties (Bibb, Chatham, Dougherty, Muskogee) according to racial disparities in COVID-19 deaths. Degree of freedom (Df), T-values, and p-values shown from Student’s T-tests shown, with false discovery rate of 10% to adjust for multiple comparisons.

A.
Representative variable	Factor	Df, T	p
Mental health provider ratio	Rural community	33.816, 4.778	<0.0001

B.
Representative variable	Factor	Df, T	p
Mental health provider ratio	Rural community	25.43, 3.144	0.004
% in frequent physical distress	Poverty	39, 2.935	0.006
PM2.5	Air pollution	21.79, 2.989	0.007
Injury death	Injury death	40, 2.644	0.012
% mammography screening	Preventive care	40, 2.499	0.017
Preventable hospitalization	Quality of care	39, 2.436	0.020
Homicide rate	Community safety	18, 2.423	0.026

Finally, we used these themes to model the greater central or peripheral pattern of racial disparities in COVID-19 deaths in counties surrounding the five Georgia cities. When greater Atlanta and central location were included in the model, greater disparity was associated with air pollution (p<0.0001; Table 3, Model 1). These relationships persisted when greater Atlanta and central location were replaced by other health-related measures. Interestingly, the new model showed better preventive care (represented by the proportion of women age 65–74 with screening mammography, Model 2) to increase the likelihood of greater racial disparity in COVID-19 deaths. This was again confirmed when deviation from the expected number of Black Georgians to die from COVID-19 in each county was used as the dependent variable, especially in counties with high injury death rates or less poverty (Model 3).

Table 3.

Factors associated with racial disparities in COVID-19 deaths for Georgia.

Greater values indicate increased likelihood of more Black COVID-19 deaths than Black population in Georgia counties.

	Model 1 Odds of greater disparity		Model 2 Odds of greater disparity		Model 3 Excess deaths, Blacks
Factors	Exp(B) (95% CI)	P	Exp(B) (95% CI)	P	B (95% CI)	p
ATL area	0.001 (0, 0.058)	0.001
Peripheral county	559 (2.2, 139590)	0.025
Air pollution, Z-score	58 (4, 807)	0.002	284 (2.8, 28626)	0.016	0.587 (0.273, 0.900)	0.001
Rural, Z-score			110 (2.1, 5781)	0.020	0.545 (0.299, 0.792)	<0.0001
Preventive care, Z-score			4.5 (0.8, 25.5)	0.086
Poverty, Z-score					−0.286 (−0.538, −0.035)	0.027
Injury death, Z-score					−0.368 (−0.726, −0.009)	0.045
Preventive care, Z-score					0.327 (0.009, 0.645)	0.044
Rural X Preventive care					−0.423 (−0.664, −0.183)	0.001
Injury death X Preventive care					0.683 (0.246, 1.121)	0.003

Discussion

Reports on hospitalized patients in Southern states suggest Black Americans to bear a disproportionate burden from COVID-19, but incomplete data on race and ethnicity from most states have so far prevented any type of population-level confirmation(Millett et al., 2020). Here we have focused on COVID-19 deaths in Georgia because of near-complete racial information in this cohort and found the greater Atlanta area to have a unique pattern of racial disparities from four other Georgia cities. Our analysis identified racial disparities to associate – at the county level – with greater air pollution, more rural community, and paradoxically better preventive care. We discuss these findings below.

Our findings suggest complex causes for the racial and geographic disparities in COVID-19 deaths. Greater racial disparities in counties with more rural communities is not surprising given the continued closure of rural hospitals (Lindrooth, Perraillon, Hardy, & Tung, 2018; O’Hanlon et al., 2019), but we did not expect better preventive care to worsen racial disparity. There are several potential explanations for this. First, mammography screening in Medicare enrollees is associated with chronic illnesses and thus exposure to the medical system (Elewonibi & Nkwonta, 2020), and a greater screening rate at the county level may reflect chronic illnesses not captured by the PC including obesity, diabetes, and physical inactivity. Screening mammography in the US is also itself associated with racial disparities (Teysir, Gegechkori, Wisnivesky, & Lin, 2019), and has actually been linked to over-utilization of healthcare resources especially in Medicare enrollees with terminal illnesses (Harris, Wilt, Qaseem, & High Value Care Task Force of the American College of, 2015; Sadigh et al., 2018). Unfortunately, screening mammography was the only variable strongly loading onto the theme of preventive care, which did not allow us to further elaborate on the latter’s nature. Thus, it would be illogical to reduce the use of screening mammography to reduce COVID-19 racial disparity. Instead, further examination of more detailed quality measures as well as other factors beyond CHR – such as rural hospitals’ affiliation with health systems (O’Hanlon et al., 2019) – strongly linked to screening mammography may identify other modifiable factors.

Air pollution was strongly associated COVID-19 racial disparities in Georgia. Air pollution has been linked to multiple chronic pulmonary, cardiovascular, as well as neurological diseases(Chen et al., 2015; Mannucci, Harari, & Franchini, 2019; Schraufnagel et al., 2019), and exposure to air pollution was estimated to contribute to approximately 5 million deaths in 2017(Cohen et al., 2017). Even before COVID-19, air pollution was found to have a disparate effect on Black and Hispanic Americans (Tessum et al., 2019). Early in the COVID-19 pandemic, SARS-CoV-2 RNA was found on PM in 20 out of 34 (54%) air samples during the Italian COVID-19 outbreak (Setti et al., 2020), and was correlated with increased COVID-19 mortality in the US and UK (Travaglio et al., 2020; Wu, Nethery, Sabath, Braun, & Dominici, 2020). This is not the case for all viral infections, as there have been conflicting findings between air pollution and influenza (du Prel et al., 2009; Somayaji et al., 2020). Interestingly, age-standardized death rates associated with air pollution actually declined in the US from 2007 to 2017, with the exception of deaths due to DM2 for ambient particulate matter. It is thus possible that a larger cohort would have allowed the detection of an interaction between air pollution and obesity, diabetes, and physical inactivity. It is further possible that lock-down and mask wearing reduce the concentration and exposure to PM-associated SARS-CoV-2. If so, residents living in areas with heavy air pollutions may preferentially benefit from these public health measures, and analysis of their impact on racial disparities in COVID-19 needs to account for air pollution levels at baseline and during lock-downs.

We used PCA to reduce the high dimensional data from CHR, and identified key themes that are measured – sometimes unexpectedly – by multiple factors. This technique is commonly used to identify latent structures within big datasets consisting of biological factors as well as social determinants of health. Because social determinants of health are hypothesized to be more correctable by policies and other human actions, it is critical to recognize that a factor nominally associated with an outcome may not be the one most likely to institute change. In our study, number of residents per dentist in a county was associated with how rural a county was, and an isolated policy increasing the number of mental health providers in rural areas is unlikely to reduce disparity. This is despite the notion that an increased number of mental health providers will be an outcome of improving rural areas. While reducing the number of screening mammography or increasing the number of mental health providers can be readily identified as red herrings, others may be more difficult to spot. Within the dentist-rural theme was also residential segregation index which was inversely associated with ruralness presumably because of greater segregation in urban areas. Yet, reducing urban residential segregation and improving rural areas are not synonymous and will likely have very different impact on communities and healthcare disparities (Caunca et al., 2020; Hayanga, Zeliadt, & Backhus, 2013). Despite air pollution’s connection to COVID-19, greater air pollution may reflect closer proximity between people due to more industrial settings or busier traffic. While observational findings in biological studies can undergo validation by experimental manipulation such as randomized controlled trials, it is not straightforward to manipulate attitudes and practices for the sake of discovery. Nevertheless, these themes at least help scientists, advocates and policymakers identify categories of social determinants for further investigation. They may additionally serve as more tangible measures of structural racism (Hahn, 2020; Hardeman, Medina, & Kozhimannil, 2016; O’Brien, Neman, Seltzer, Evans, & Venkataramani, 2020; Waite, Sawyer, & Waite, 2020) for longitudinal trend monitoring, with the caveats that the latent variables underlying each theme may differ between states and between communities within the same state.

Conclusion

Blacks were over-represented in COVID-19 deaths for most counties surrounding five major Georgia cities. These disparities were worse in communities with worse air pollution (which can harbor the coronavirus) and more residents living in rural areas. Public health practices such as mask wearing and lockdowns may benefit Black Georgians by reducing the pollutants they inhale, but longer term goals need to address reducing access to quality medical care in rural regions.

Table 1.

Characteristics of counties included in the current analysis.

Information from Georgia DPH and CHR as of 6/12/2020. Counties with no COVID-19 deaths but are part of the peripheral counties are shown in gray.

	Atlanta, GA	Albany, GA	Columbus, GA	Macon, GA	Savannah, GA
Central county	Fulton	Dougherty	Muscogee	Bibb	Chatham
County population, thousands	1,050.1	91.2	194.2	153.1	289.2
% rural	1.1%	14.0%	3.0%	14.4%	4.5%
Age % <18 % >65	21.8% 11.7%	23.3% 15.4%	24.8% 13.4%	24.6% 15.6%	21.2% 15.4%
Female	51.6%	54.1%	51.4%	53.1%	51.9%
Race Asian Black Hispanic White Other	78,463 (7%) 457,994 (44%) 76,333 (7%) 416,624 (40%) 20,700 (2%)	917 (1%) 64,103 (70%) 2,672 (3%) 22,299 (24%) 1,252 (1%)	5,426 (3%) 90,133 (46%) 14,888 (8%) 77,758 (40%) 5,955 (3%)	3,367 (2%) 84,136 (55%) 5,216 (3%) 57,923 (38%) 2,453 (2%)	8,447 (3%) 116,022 (40%) 18,968 (7%) 139,361 (48%) 6,397 (2%)
No. COVID-19 cases No. COVID-19 deaths (%) No. Black COVID-19 deaths (%)	4,989 273 (5.5%) 202 (74%)	1,820 150 (8.2%) 118 (79%)	904 23 (2.5%) 12 (52%)	516 34 (6.6%) 20 (59%)	562 29 (5.2%) 10 (34%)
Surrounding counties	Carroll Cherokee Clayton Cobb Coweta DeKalb Douglas Fayette Forsyth Gwinnett	Baker Calhoun Colquitt Lee Mitchell Terrell Worth	Chattahoochee Harris Macon Marion Schley Stewart Sumter Talbot Taylor Webster	Crawford Houston Jones Monroe Peach Twiggs Wilkinson	Bryan Bulloch Effingham Evans Liberty Long McIntosh
County population, thousands (SD)	374.4 (312.3)	19.4 (15.0)	13.2 (11.2)	38.3 (52.4)	40.4 (26.8)
% rural	13.8% (14.6%)	67.3% (24.4%)	79.4% (29.1%)	70.9% (35.2%)	58.2% (19.3%)
Age % <18 (SD) % >65 (SD)	25.0% (1.8%) 12.7% (2.5%)	22.4% (3.2%) 17.3% (3.3%)	20.7% (2.5%) 18.1% (5.7%)	21.8% (2.1%) 18.0% (3.1%)	24.8% (4.8%) 13.6% (6.2%)
Female (SD)	51.6% (0.9%)	49.0% (4.3%)	49.1% (5.5%)	51.1% (0.9%)	50.4% (0.8%)
Race Asian (SD) Black (SD) Hispanic (SD) White (SD) Other (SD)	5.6% (4.6%) 29.7% (21.0%) 10.9% (4.4%) 51.7% (22.5%) 2.1% (0.2%)	1.1% (0.8%) 40.6% (16.6%) 6.2% (6.1%) 50.8% (14.7%) 1.3% (0.1%)	1.2% (0.8%) 36.9% (16.4%) 6.0% (4.1%) 54.1% (14.6%) 1.9% (1.1%)	1.1% (1.0%) 31.7% (9.3%) 3.9% (2.4%) 61.4% (10.8%) 1.9% (0.4%)	1.4% (0.6%) 26.9% (10.2%) 7.8% (4.2%) 61.2% (12.9%) 2.6% (1.4%)
No. COVID-19 cases (SD) No. COVID-19 deaths (SD) Case fatality rate, % (SD) % Black COVID-19 deaths (SD)	1,726 (1,715) 67.9 (69.6) 4.0% (1.5%) 39.5% (24.2%)	291 (190) 18.6 (11.4) 7.1% (3.1%) 70.6% (15.8%)	119 (147) 7.0 (14.6) 5.0% (3.0%) 61.4% (40.1%)	112 (135) 6.9 (7.5) 5.6% (4.3%) 72.4% (27.4%)	51 (35) 1.5 (1.7) 3.0% (2.7%) 21.9% (36.9%)

Funding Statement

This work was supported by National Institutes of Health grants R01 AG 054046 and R01 AG054991.