Association of Race With Mortality Among Patients Hospitalized With Coronavirus Disease 2019 (COVID-19) at 92 US Hospitals

IMPORTANCE While current reports suggest that a disproportionate share of US coronavirus disease 2019 (COVID-19) cases and deaths are among Black residents, little information is available regarding how race is associated with in-hospital mortality. OBJECTIVE To evaluate the association of race, adjusting for sociodemographic and clinical factors, on all-cause, in-hospital mortality for patients with COVID-19. DESIGN, SETTING, AND PARTICIPANTS This cohort study included 11 210 adult patients (age 18 years) hospitalized with confirmed severe acute respiratory coronavirus 2 (SARS-CoV-2) between February 19, 2020, and May 31, 2020, in 92 hospitals in 12 states: Alabama (6 hospitals), Maryland (1 hospital), Florida (5 hospitals), Illinois (8 hospitals), Indiana (14 hospitals), Kansas (4 hospitals), Michigan (13 hospitals), New York (2 hospitals), Oklahoma (6 hospitals), Tennessee (4 hospitals), Texas (11 hospitals), and Wisconsin (18 hospitals). EXPOSURES Confirmed SARS-CoV-2 infection by positive result on polymerase chain reaction testing of a nasopharyngeal sample. MAIN OUTCOMES AND MEASURES Death during hospitalization was examined overall and by race. Race was self-reported and categorized as Black, White, and other or missing. Cox proportional hazards regression with mixed effects was used to evaluate associations between all-cause in-hospital mortality and patient characteristics while accounting for the random effects of hospital on the outcome. RESULTS Of 11 210 patients with confirmed COVID-19 presenting to hospitals, 4180 (37.3%) were Black patients and 5583 (49.8%) were men. The median (interquartile range) age was 61 (46 to 74) years. Compared with White patients, Black patients were younger (median [interquartile range] age, 66 [50 to 80] years vs 61 [46 to 72] years), were more likely to be women (2259 [49.0%] vs 2293 [54.9%]), were more likely to have Medicaid insurance (611 [13.3%] vs 1031 [24.7%]), and had higher median (interquartile range) scores on the Neighborhood Deprivation Index (−0.11 [−0.70 to 0.56] vs 0.82 [0.08 to 1.76]) and the Elixhauser Comorbidity Index (21 [0 to 44] vs 22 [0 to 46]). All-cause in-hospital mortality among hospitalized White and Black patients was 23.1% (724 of 3218) and 19.2% (540 of 2812), respectively. After adjustment for age, sex, insurance, comorbidities, neighborhood deprivation, and site of care, there was no statistically significant difference in risk of mortality between Black and White patients (hazard ratio, 0.93; 95% CI, 0.80 to 1.09). CONCLUSIONS AND RELEVANCE Although current reports suggest that Black patients represent a disproportionate share of COVID-19 infections and death in the United States, in this study, mortality (continued) Key Points Question Is race associated with mortality among patients hospitalized with coronavirus disease 2019 (COVID-19) in the United States? Findings In this cohort study of 11 210 individuals with COVID-19 presenting for care at 92 hospitals across 12 states, there was no difference in all-cause, in-hospital mortality between White and Black patients after adjusting for age, sex, insurance status, comorbidity, neighborhood deprivation, and site of care. Meaning In this study, race was not independently associated with in-hospital mortality after adjusting for differences in sociodemographic and clinical factors. + Invited Commentary Author affiliations and article information are listed at the end of this article. Open Access. This is an open access article distributed under the terms of the CC-BY License. JAMA Network Open. 2020;3(8):e2018039. doi:10.1001/jamanetworkopen.2020.18039 (Reprinted) August 18, 2020 1/9 Abstract (continued)continued) for those able to access hospital care did not differ between Black and White patients after adjusting for sociodemographic factors and comorbidities. JAMA Network Open. 2020;3(8):e2018039. doi:10.1001/jamanetworkopen.2020.18039


Introduction
As coronavirus disease 2019 (COVID-19) continues to spread across the United States, understanding how race is associated with outcomes will be important to clinicians, health systems, and public officials responding to this pandemic. 1,2 Current reports suggest that a disproportionate share of US COVID-19 cases are among Black residents (21%) compared with their proportion of the US population (13%). 3 However, much of our national-level data relies on aggregated local, state, and territorial reports to the US Centers for Disease Control and Prevention (CDC), which have high proportions (55%) of missing race/ethnicity data. 4 In 20 of 31 states reporting COVID-19 cases by race/ethnicity, Black patients accounted for a greater share of confirmed cases compared with their share of the total population. 4,5 Additionally, Black patients represented more than half of all COVID-19 deaths in Alabama (52%), Georgia (51%), Louisiana (59%), Mississippi (66%), and the District of Columbia (75%). 4,5 To date, US studies evaluating the association of race with COVID-19 outcomes have mostly been limited to summary statistics, with large amounts of missing data or localization to specific states. [3][4][5][6][7][8][9][10][11][12] We performed analyses on a large cohort of patients with COVID-19 to better understand the association between race and mortality, alone and in combination with age, sex, insurance status, comorbidity, and social factors, among patients who accessed hospital care for their disease.

Methods
Adults (age Ն18 years) with confirmed severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection (positive result by polymerase chain reaction testing of a nasopharyngeal sample) presenting to Ascension hospitals between February 19, 2020, and May 31, 2020, were included. To fully capture hospital outcomes, patients admitted during this time period were observed through June 25, 2020. Hospitals were located in 12 states: Alabama (6 hospitals), Maryland (1 hospital), Florida (5 hospitals), Illinois (8 hospitals), Indiana (14 hospitals), Kansas (4 hospitals), Michigan (13 hospitals), New York (2 hospitals), Oklahoma (6 hospitals), Tennessee (4 hospitals), Texas (11 hospitals), and Wisconsin (18 hospitals). As a system, standard protocols were established and implemented for COVID-19 screening, testing, and management across the sites of care. The Ascension Seton institutional review board approved the study protocol as exempt and granted a waiver of consent because this retrospective study was based on secondary use of data arising from routine care. This study followed the Strengthening the Reporting of Observational addition, ICD-10 diagnosis codes were used to assess the following coexisting conditions: asthma, cancer, chronic kidney disease, chronic liver disease, chronic obstructive pulmonary disease, congestive heart failure, coronary artery disease, diabetes, hypertension, obesity, and solid organ transplantation. Diagnosis codes used to calculate the ECI and examine the presence of coexisting conditions were drawn from hospital billing data for the patient encounters included in this analysis.
If available, first documented temperature, respiratory rate, and oxygen saturation level were extracted from the electronic health record. Using each patient's home address, we assigned the neighborhood deprivation index (NDI) for their zip code. The NDI is a composite of social and material deprivation composed of 8 variables collected in the American Community Survey focused on poverty, employment, education, and housing. 15 The NDI ranged from −1.80 to 3.23 in our sample, with higher numbers indicating greater deprivation. We described summary statistics for patient sociodemographic characteristics, health care utilization (intensive care unit [ICU] stay, use of mechanical ventilation), and all-cause in-hospital mortality (indicated by use of hospital discharge status code 20, expired).

Statistical Analysis
Cox proportional hazards regression with mixed effects was used to evaluate associations between patient characteristics and time to all-cause, in-hospital mortality by calculating hazard ratios (HRs) with 95% CIs. These models account for cluster-specific random effects that may result in differing baseline hazard functions between hospitals. We used 3 models to assess the association of race with mortality; all models included the random effects term for hospital. Model 1 estimated HRs for patient characteristics, unadjusted for other factors. Model 2 included a priori patient characteristics of interest (race, age, sex, insurance, and ECI and NDI scores). Model 3 adjusted for the covariates in model 2 and specific chronic conditions if they were significant in univariable analyses at P < .05. The proportional hazards assumption for the Cox models was assessed and confirmed graphically.
Sensitivity analyses were performed to investigate potential residual confounding by modeling age as a categorical variable rather than as a continuous variable. Two-sided testing was used, with P < .05 considered statistically significant. Statistical analyses were performed using R version 3.6.2 (R Project for Statistical Computing).

Results
Of 11  mortality than those without these conditions. Adjusting for patient sociodemographic and clinical factors, race was not significantly associated with an increased risk of death (HR, 0.93; 95% CI, 0.83-1.09) (Figure 2). Sensitivity analysis modeling categorical age groups vs continuous age revealed a nonlinear association between age and mortality, with older age groups having a higher risk of death than a corresponding risk calculated from the continuous age model estimate. Additionally, both approaches yielded nearly identical estimates for all covariates.

Discussion
In our cohort of 92 hospitals across 12 states, there was no statistically significant difference in all-cause, in-hospital mortality between White and Black patients after adjusting for other factors.
Overall all-cause, in-hospital mortality was 20.3%; it was 38.1% among patients receiving mechanical ventilation, similar to findings from other US studies. 3,7,10,11,16 Consistent with prior research, older age and male sex were significantly associated with risk of death. 4,6,7,10,11,17 Current data indicate that Black residents represent a disproportionate share of US COVID-19 cases and deaths. [3][4][5] Living conditions (eg, residence in densely populated areas; multigeneration households; the presence of racial segregation; overrepresentation in jails, prisons, and detention centers), work circumstances (eg, higher rates of essential workers; lack of paid sick leave), and underlying health conditions (eg, lack of health insurance; higher rates of certain comorbidities; chronic and toxic stress associated with stigma and systemic inequalities) may help explain the disproportionate burden of illness and death among patients who belong to minority groups. 9,18 Our findings confirm data from Louisiana (a cohort of patients seeking care at a New Orleans-based integrated delivery health system) 10 and Georgia (a convenience sample of hospitalized adult patients in metropolitan Atlanta and southern Georgia), 11 which reported no difference in mortality by race among hospitalized patients with COVID-19. Taken together, these findings suggest that while Black US residents may be at higher risk of contracting COVID-19 and represent a disproportionate share of COVID-19 death, mortality for those able to access hospital care does not differ from White patients.    In our sample, older age was the strongest independent risk factor for death, with White patients a median of 6 years older than Black patients. This age difference between patients admitted to the hospital with COVID-19 is consistent with other studies. 3,10,11 Similar to prior reports, we noted higher rates of diabetes, hypertension, chronic kidney disease, and obesity among Black patients. 10 respectively). This is consistent with reports from Georgia but different from Louisiana data, which showed higher ICU and invasive mechanical ventilation among Black patients. 10,11 Limitations This study has several limitations. First, our study focused on individuals able to access hospital care.
We did not observe patients before or after discharge. Mortality may vary when accounting for death prior to and after hospitalization. Second, collection of ethnicity data varied by hospital. Therefore, we were unable to examine the association of ethnicity with mortality. Third, neighborhood deprivation was measured at the zip code level and may not reflect individual factors. Similarly, ECI is based on ICD-10 diagnosis codes, which may not fully reflect patients' comorbidities. It is possible that documentation of diagnosis codes for some patients was incomplete. Fourth, while our hospitals are located in diverse settings across multiple states, our experience may not be representative of other hospitals.

Conclusions
In this analysis of patients with COVID-19, higher risk of all-cause in-hospital mortality was associated with older age, male sex, Medicare insurance, coexisting chronic kidney disease, and coronary artery disease. No statistically significant difference by race was observed. Additional studies examining COVID-19 mortality by race, accounting for death prior to and after hospitalization, are needed.