Article Text

Pandemic phase-related racial and ethnic disparities in COVID-19 positivity and outcomes among patients presenting to emergency departments during the first two pandemic waves in the USA
Free
  1. Shaveta Khosla1,
  2. Marina Del Rios1,2,
  3. Makini Chisolm-Straker3,
  4. Saadiyah Bilal3,
  5. Timothy B Jang4,
  6. Hao Wang5,
  7. Molly Hartley6,
  8. George T Loo3,
  9. James P d'Etienne5,
  10. Craig D Newgard7,
  11. D Mark Courtney8,
  12. Esther K Choo7,
  13. Michelle P Lin3,9,
  14. Jeffrey A Kline10
  1. 1 Emergency Medicine, University of Illinois Chicago, Chicago, Illinois, USA
  2. 2 Emergency Medicine, University of Iowa, Iowa City, Iowa, USA
  3. 3 Emergency Medicine, Icahn School of Medicine at Mount Sinai, New York, New York, USA
  4. 4 Harbor-UCLA Medical Center, Emergency Medicine, David Geffen School of Medicine at UCLA, Torrance, California, USA
  5. 5 Emergency Medicine, John Peter Smith Health Network, Fort Worth, Texas, USA
  6. 6 Portsmouth Regional Hospital, Portsmouth, New Hampshire, USA
  7. 7 Emergency Medicine, Oregon Health & Science University, Portland, Oregon, USA
  8. 8 Emergency Medicine, UT Southwestern, Dallas, Texas, USA
  9. 9 Emergency Medicine, Stanford University, Stanford, California, USA
  10. 10 Emergency Medicine, Wayne State University, Detroit, Michigan, USA
  1. Correspondence to Dr Shaveta Khosla, Emergency Medicine, University of Illinois Chicago, Chicago, Illinois, USA; skhosl2{at}uic.edu

Abstract

Background In many countries including the USA, the UK and Canada, the impact of COVID-19 on people of colour has been disproportionately high but examination of disparities in patients presenting to ED has been limited. We assessed racial and ethnic differences in COVID-19 positivity and outcomes in patients presenting to EDs in the USA, and the effect of the phase of the pandemic on these outcomes.

Methods This is a retrospective cohort study of adult patients tested for COVID-19 during, or 14 days prior to, the index ED visit in 2020. Data were obtained from the National Registry of Suspected COVID-19 in Emergency Care network which has data from 155 EDs across 27 US states. Hierarchical models were used to account for clustering by hospital. The outcomes included COVID-19 diagnosis, hospitalisation at index visit, subsequent hospitalisation within 30 days and 30-day mortality. We further stratified the analysis by time period (early phase: March–June 2020; late phase: July–September 2020).

Results Of the 26 111 adult patients, 38% were non-Hispanic White (NHW), 29% Black, 20% Hispanic/Latino, 3% Asian and 10% all others; half were female. The median age was 56 years (IQR 40–69), and 53% were diagnosed with COVID-19; of those, 59% were hospitalised at index visit. Of those discharged from ED, 47% had a subsequent hospitalisation in 30 days. Hispanic/Latino patients had twice (adjusted OR (aOR) 2.3; 95% CI 1.8 to 3.0) the odds of COVID-19 diagnosis than NHW patients, after adjusting for age, sex and comorbidities. Black, Asian and other minority groups also had higher odds of being diagnosed (compared with NHW patients). On stratification, this association was observed in both phases for Hispanic/Latino patients. Hispanic/Latino patients had lower odds of hospitalisation at index visit, but when stratified, this effect was only observed in early phase. Subsequent hospitalisation was more likely in Asian patients (aOR 3.1; 95% CI 1.1 to 8.7) in comparison with NHW patients. Subsequent ED visit was more likely in Blacks and Hispanic/Latino patients in late phase.

Conclusion We found significant differences in ED outcomes that are not explained by comorbidity burden. The gap decreased but persisted during the later phase in 2020.

  • emergency department
  • COVID-19
  • hospitalisations

Data availability statement

Data are available on reasonable request.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

WHAT IS ALREADY KNOWN ON THIS TOPIC

  • There is a higher burden of COVID-19 among people of colour in the general population.

WHAT THIS STUDY ADDS

  • There are racial/ethnic differences in COVID-19 positivity and outcomes in the ED patient population, and the effect varies by the phase of pandemic.

  • Asian patients had higher likelihood of subsequent hospitalisation compared with White patients.

  • Hispanic/Latino patients had higher likelihood of COVID-19 diagnosis (irrespective of the phase of the pandemic).

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

  • Future research should assess the mechanisms by which pandemics exacerbate health inequities in order to prioritise policies and public health programming aimed at prevention, early diagnosis and treatment parity.

Introduction

In the USA, the COVID-19, caused by SARS-CoV-2, has disproportionately impacted people of colour.1 2 Other countries, including the UK and Canada, have also observed COVID-19 disparities that have impacted racial and ethnic minority groups.3–5 Multiple US-based studies have continued to highlight disparities with respect to COVID-19 incidence and clinical outcomes,1 6 7 and these disparities are persisting over time.2 6 The Centers for Disease Control and Prevention estimates that Black, Hispanic/Latino and American Indian/Alaska Native patients have twice the age-adjusted risk of hospitalisation and death from COVID-19 compared with non-Hispanic White (NHW) patients while Asian patients have a lower comparative risk.8 American Indian/Alaska Native and Hispanic/Latino also have a 1.5 times elevated (age-adjusted) risk of COVID-19 infection compared with NHW patients.8

While the prior studies have documented higher rates of COVID-19 infection and hospitalisation among people of colour, the examination of disparities among patients presenting to the ED for COVID-19 has been limited geographically or by sample size. The ED patient population tends to be somewhat different from other clinical settings as EDs often serve as a safety net for patients that are uninsured or those that lack access to primary care.9 Some racial and ethnic groups also prefer to access care through ED.9–11 Additionally, studies have not compared the phases within 2020, which is important given the rapid changes that were occurring in 2020. The decision to admit patients to the hospital after ED evaluation is typically dependent on illness severity; however, capacity constraints experienced during the COVID-19 pandemic in 2020 led to changes in usual practice such as discharging patients from the ED who would normally qualify for hospital admission.12 13 Individual and structural racism can influence likelihood of admission even in the absence of a pandemic and impact hospital admission rates and clinical outcomes.14 15 Early in the pandemic, when testing was reserved for those with higher probability of disease, COVID testing was not equally accessible,16 and may have been allocated selectively. There is a need for more comprehensive investigation of racial and ethnic differences in COVID-19 incidence and clinical outcomes, including examining the impact of ED disposition, delayed hospitalisations and how these factors impact subsequent hospitalisation, ED visits and clinical outcomes by the phase of pandemic in patients accessing care in ED. In this study, we assessed racial and ethnic differences in COVID-19 positivity, and COVID-19 outcomes in patients presenting to the ED and the effect of the phase of the pandemic on these outcomes, using a national multicentre registry based in the USA.

Methods

This is a retrospective cohort study of adults (aged ≥18 years) presenting to the US EDs with COVID-19-like illnesses in 2020. In this cohort, ‘COVID-19-like illness’ was defined as having a triage chief complaint concerning for COVID-19 infection such as fever, sore throat, cough or shortness of breath, and the patient was tested for COVID-19. Data were obtained from the US-based National Registry of Suspected COVID-19 in Emergency Care (RECOVER) network, a large ED-based COVID-19 registry with patient data from 155 EDs across 27 US states. The registry includes ED patients with a SARS-CoV-2 test during, or 14 days prior to, the index visit, enrolled between March and September 2020 with the intent to enrol eligible patients consecutively. Briefly, the registry includes data abstracted from electronic medical records (EMR) for the patients’ index visit, that is, the first ED visit within 14 days of SARS-CoV-2 testing. Patients were eligible for enrolment if they had a molecular diagnostic test ordered and performed in the ED with SARS-CoV-2 infection or COVID-19 suspected. Patients were excluded if there was a lack of reasonable probability of COVID-19 infection (such as when the test was done for administrative/policy purposes and was not based on clinical suspicion of COVID-19 infection). Outcomes were recorded up to 30 days after the index visit with all follow-ups being done through EMR abstractions at each site. More details on the registry can be found in the protocol methodology published previously.17

Data were entered into Research Electronic Data Capture (REDCap). To maintain quality, REDCap forms were inspected centrally, and verification queries were generated and resolved with the site. For certain variables, missing data were assessed for monotonicity and the missing values were replaced through multiple imputation method. More details on the imputation methods are included in the study by Kline et al.18 For this study, the variables that had imputed values include: age, sex, RR and oxygen saturation. Of the total of 27 051 presentations of COVID-19-like illness included in the RECOVER dataset, age was imputed for 0.24%, sex 0.06%, RR 1.2% and lowest oxygen saturation in the ED for 3.9% patients. We excluded 940 paediatric cases from our analysis.

Self-identified race and ethnicity, as reported in EMR, was categorised as NHW (ie, identified race as White, and ethnicity as not Hispanic or Latino), Black, Hispanic/Latino (ie, identified ethnicity as Hispanic or Latino, irrespective of race except when multiple race categories were checked), Asian, Native Hawaiian/Pacific Islander (Native Hawaiian or Other Pacific Islander), American Indian/Alaska Native and unknown/mixed/other. Unknown/Mixed/Other was selected when ethnicity was not labelled as ‘Hispanic or Latino’ (ie, either ethnicity was ‘not Hispanic or Latino’, or it was ‘unknown’ or was missing) and one of the conditions were met: (1) race was marked as ‘unknown/other, (2) none of the race categories were checked or (3) if more than one race was marked.

The outcomes assessed included: (1) SARS-CoV-2 infection: a patient was considered positive for SARS-CoV-2 infection if either a molecular diagnostic test from a swab or serological IgM or IgG antibody test within 30 days was positive. A patient was considered negative for SARS-CoV-2 in the absence of positive test results or clinical diagnosis of COVID-19 within 30 days. Within the sample that was SARS-CoV-2 positive, we further assessed the following outcomes: (2) admission at index ED visit (missing for 11 patients), (3) subsequent admission within 30 days after ED discharge for patients who returned to ED (missing for one patient), (4) subsequent ED visit within 30 days of ED discharge (missing for seven patients), (5) 30-day mortality (patients documented to have died within 30 days of index ED visit were marked ‘yes’, and all others were marked ‘no’; therefore, if a patient’s death was not reported in EMR, it would not be marked ‘yes’). The covariates included sex, age, healthcare insurance and comorbidity (obesity, diabetes, hypertension, heart failure, chronic obstructive pulmonary disease, prior venous thromboembolism).19–21 We also examined clinical severity indicators including RR and lowest oxygen saturation while in the ED.

Descriptive statistics were generated to characterise the study population. Bivariate analyses were conducted to assess differences in characteristics by race and ethnicity using χ2 test and Fisher’s exact test for categorical variables, and analysis of variance and Kruskal-Wallis test for continuous variables. Unless categorised under unknown, missing values were excluded from the respective final models. To address potential clustering by hospital, we ran hierarchical models using generalised estimating equations with logit link estimates (GENMOD Procedure) to assess the association of each of the outcomes with race and ethnicity, after adjusting for (a) sex, age, (b) sex, age and clinical variables, (c) sex, age and comorbidities. An exchangeable correlation structure was used. There were in all 120 clusters and the average number of patients in a cluster was 218. In addition to the hierarchical models, multivariable logistic models were also run and those results were presented separately. Akaike information criterion and Hosmer-Lemeshow test were used to assess the fit of models. Furthermore, to assess the effect of pandemic phase, we also examined the association of race/ethnicity with each of the outcomes when stratified by time period (early phase: March–June 2020, late phase: July–September 2020). This stratification was planned a priori. Race/Ethnicity and phase of pandemic interaction term p values were also assessed. Due to the geography of participating sites, the sample was smaller for American Indian/Alaska Native and Native Hawaiians/Pacific Islander when assessing certain outcomes, and we excluded these two groups from those analyses. We used SAS V.9.4 (Statistical Analysis Software; SAS Institute, Cary, North Carolina, USA) for the analyses.

Patient and public involvement

Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

Results

A total of 27 051 patients were documented in the US-based registry as of December 2020. Exclusion of paediatric cases left a resulting sample of 26 111 adult patients presenting to the ED with COVID-19-like illness. Of these, 38% were NHW patients, 29% were Black, 20% were Hispanic/Latino, 3% were Asian, 0.4% were American Indian/Alaska Native, 0.3% were Native Hawaiian/Pacific Islander and 9.6% were unknown/mixed/pthers. Half of the cohort (50.4%) was female, and the mean age was 55 years (SD=19).

Hispanic/Latinos and American Indian/Alaska Native patients were younger than other patients (table 1). Hispanic/Latino patients were most likely to be uninsured, while NHW patients had the lowest likelihood of being uninsured. Obesity diagnosis was least common in Asian patients. Diabetes was most common among Native Hawaiians/Pacific Islander.

Table 1

Characteristics of the sample by racial and ethnic group (n=26 111)

Of the total sample, 53% (n=13 810) were diagnosed with SARS-CoV-2. Distribution of COVID-19-related variables by race/ethnicity is presented in table 2. Hispanic/Latino patients had the highest likelihood of being positive for SARS-CoV-2 (74% vs 31% in NHW, p<0.001). Black, Asian and unknown/mixed/other group also had a high likelihood of SARS-CoV-2 diagnosis.

Table 2

COVID-19 outcomes

When the analysis was restricted to the cohort that had SARS-CoV-2 infection, Asian patients were more likely to have higher RRs (>28 per minute) and had the highest proportion of measured low oxygen saturation (ie, suboptimal oxygenation with a saturation of 92 or below) in the ED. NHWs and Asians were most likely to be admitted, while American Indian/Alaska Native patients had the lowest admission rate. Among those that had been discharged from ED at the index visit, subsequent hospitalisation was most common in Asian patients. Subsequent ED visit was most common in American Indian/Alaskan Native patients. Mortality within 30 days of index visit was highest in unknown/mixed/other racial group, followed by NHW and Asian groups.

In adjusted analyses (clustered by hospital) using the full cohort of patients, Hispanic/Latino patients (adjusted OR (aOR) 2.3; 95% CI 1.8 to 3.0) and Black patients (aOR 1.7; 95% CI 1.4 to 2.0) had higher odds of being diagnosed with SARS-CoV-2 infection than NHW patients (table 3). Similarly, Asian, Native Hawaiian/Pacific Islander, American Indian/Alaska Native and unknown/mixed/other patients had higher odds of being positive for SARS-CoV-2 (compared with NHW patients). The results of a comparable model that is not clustered by hospital are presented in online supplemental table 1. When not accounting for clustering, the association tended to be stronger for certain outcomes. When stratified by the pandemic phase in 2020 (p value for interaction term <0.0001), the effect was stronger in the early phase with Hispanic/Latino (aOR 2.2; 95% CI 1.7 to 3.0), Black (aOR 1.7; 95% CI 1.5 to 2.1), Asian (aOR 1.6; 95% CI 1.3 to 1.9) and unknown/mixed/other (aOR 1.6; 95% CI 1.3 to 1.8) patients having higher odds of being diagnosed with SARS-CoV-2 compared with NHW patients, after adjusting for age, sex and comorbidities. In the late phase, the association persisted only in Hispanic/Latino patients (table 4). Comparable stratified models when not clustered by hospital are presented in online supplemental table 2.

Supplemental material

Table 3

Crude and adjusted results from hierarchical models for the association of race and ethnicity with key outcomes, accounting for clustering by hospital

Table 4

Results stratified by the phase of the pandemic, accounting for clustering by hospital and adjusted for age groups, sex and comorbidities

In analyses limited to the cohort of SARS-CoV-2 cases, Hispanic/Latino patients (aOR 0.8; 95% CI 0.7 to 1.0) had lower odds of being admitted to the hospital at the index visit when compared with NHW, after adjusting for age, sex and comorbidities. When stratified by the phase of pandemic, this lower likelihood of being admitted to the hospital at the index visit among Hispanic/Latino patients was only present in the early phase. Among those discharged from the ED at the time of index visit, subsequent hospitalisation within 30 days was more likely in Asian patients (aOR 3.1; 95% CI 1.1 to 8.7) in comparison with NHW. The p value for race/ethnicity and pandemic phase interaction term in the models predicting admission at index visit and subsequent ED visit was 0.005 and 0.48, respectively. When subsequent ED visits were examined among those that were discharged from ED at the time of index visit, the association was non-significant for all groups of colour compared with NHW patients, except unknown/mixed/other group when adjusted for age, sex and comorbidities. However, when stratified by phase of pandemic, the later phase of the pandemic observed higher likelihood of subsequent ED visit among Black and Hispanic/Latino patients compared with NHW patients. When compared with NHW patients, mortality within 30 days of index visit was lower in Black and Hispanic/Latino patients. This pattern was present in the early phase of the pandemic but not in the later phase.

Discussion

In this US cohort of ED patients presenting with COVID-19-like illness, we found higher adjusted odds of testing positive for COVID-19 infection among minoritised groups including Hispanic/Latino, Black and Asian patients, but lower adjusted odds of hospitalisation among Hispanic/Latino patients compared with NHW patients, while subsequent hospitalisation after ED discharge was highest among Asian patients. These findings persisted even after controlling for comorbidities in addition to age and sex. Even when clinical severity of infection was accounted for, Hispanic/Latinos had lower odds of admission at index visit, which may suggest an actual disparity which cannot be attributed to Hispanic/Latino patients being less sick on ED presentation. The highest odds of being SARS-CoV-2 positive in ED was among Hispanic/Latino patients. These findings are consistent with those reported by Magesh et al 22 and Serrano et al.23 Furthermore, our results are coherent with the worse postacute sequelae of COVID-19 symptoms and conditions observed in racial/ethnic minority groups.24 The use of only EMR data substantially limited our ability to examine mortality and that may explain why our finding of lower likelihood of 30-day mortality among Hispanic/Latino and Black patients conflicts with a report published in 2020 that found a higher mortality burden in Hispanic and non-Hispanic Black after age-standardisation.25

In extant literature, the phases of the pandemic have not been explored much with respect to ED patients. Although one study based in Canada included waves of the pandemic in the analysis, the focus was on comparing the long-term (physical and mental health) quality of life outcomes of SARS-CoV-2-positive and SARS-CoV-2-negative cases.26 While the COVID-19 diagnostic testing strategy, public perception of the severity of the infection and treatment guidelines continued to evolve through 2020, we assessed the impact of this by stratifying for the pandemic phase during the first two waves of the pandemic in 2020. The early phase showed a much stronger association of race/ethnicity with ED visits. The most striking effect was that Black and Asian patients and patients with unknown/mixed/other race and ethnicity had higher likelihood of being diagnosed with SARS-CoV-2 infection (compared with NHW) in the early phase (but not in the late phase). However, for Hispanic/Latino patients, the higher likelihood of being diagnosed with SARS-CoV-2 infection was observed in both phases. On stratification, the sample size was small for certain groups and that may have led to the wide CIs noticed for some associations.

The sources of the inequities in SARS-CoV-2 infection rates and subsequent outcomes among patients presenting with COVID-19-like illness in the ED are complex and may be a result of structural inequities.27–29 Black, Latino, Asian and Native American people are more likely to be employed in essential areas of the economy.30 31 In the early part of the pandemic when many businesses were closed and people were encouraged to work from home, essential workers in service industries, manufacturing and healthcare had to work in person.32 33 Essential workers are also more likely to be exposed to COVID-19.31 Many also live in multifamily homes,34 where isolation posthigh-risk exposure is not possible. Unequal access to health insurance compounds unequal access to primary and preventive care services which in turn increases the burden of pre-existing comorbidity.35 It is important to note that early in the pandemic, when health systems experienced testing scarcity,16 structural racism (which leads to persistence of inequitable access to high-quality healthcare)14 15 may have contributed to unequal testing rates, which would bias our results to an underestimate of test positivity among people of colour.

Certain changes were observed from early to late phase with some associations diminishing, while other patterns reversed or trended towards reversing, with race/ethnicity and phase of pandemic interaction term p values being highly significant in model predicting SARS-CoV-2 infection, and model predicting admission at index visit. Potential explanations warranting further study include standardisation of care later in the pandemic, including more specific parameters for ED disposition, thereby, reducing potential for subjective and bias-laden disposition decisions; and increased utilisation of therapies like dexamethasone and high flow nasal cannula in favour of intubation that led to an overall reduction in mortality. Moreover, increased availability of outpatient testing for SARS-CoV-2 may have directed less symptomatic patients away from the ED, thereby narrowing the disparities in ED positivity rates.

Limitations

We used all-cause mortality, which may not accurately represent COVID-19-related mortality. Mortality data may have been incomplete as we used EMR at participating institutions and did not use death certificate data or linked EMR from other institutions; thereby, information may have been missed for patients that received care at more than one healthcare facility or who died outside of a healthcare facility. For a segment of our cohort, race/ethnicity data were missing or unknown. This group had high levels of poor outcomes and needs to be better defined and examined in future studies. Moreover, we had patients who identified as more than one race, reflecting the reality of US demographics. Instead of excluding these cases completely, we included them in the analysis as a separate category to assess the patterns observed among those that identified as unknown/other/mixed race. We did not have information on education or income level of the patients; therefore, we could not assess the effect of these factors. As the registry was limited to ED patients who had received a COVID-19 diagnostic test, our sample is likely to be biased towards those that had symptoms severe enough to go to the ED, who had the means to visit the ED or felt safe enough to seek ED care. Additionally, many hospitals restricted testing early in the pandemic due to insufficient availability, therefore, testing bias cannot be ruled out. False negatives related to the timing of the test can also bias the results such that groups that presented later in the illness may show a higher positivity rate. Therefore, this sample may not be representative of mild or asymptomatic cases, people who lacked transportation or people who had concerns about visiting a healthcare institution.

The strengths of this study are the use of a large, national (US) dataset which allowed for comprehensive analysis of COVID-19 by race and ethnicity. To our knowledge this is the first study in the US that has compared the early phase of the pandemic to late phase in 2020, to examine the impact of the pandemic on different racial and ethnic groups.

Conclusion

We found significant differences in ED outcomes, including confirmed COVID-19 infection, admission at index visit and subsequent hospitalisation, by race and ethnicity in this retrospective cohort study. These inequities in the ED patients are not explained by comorbidity burden. The gap decreased for some outcomes but persisted (or worsened for others) during the late phase of the cohort study period. More research is needed to understand the mechanisms by which pandemics exacerbate health inequities. Such data can better inform policies and public health programming aimed at prevention, early diagnosis and treatment parity.

Abstract translation

This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Data availability statement

Data are available on reasonable request.

Ethics statements

Patient consent for publication

Ethics approval

The protocol for the registry was reviewed and approved by the Institutional Review Boards at all participating sites, and was approved under waiver of authorisation for participation in research and waiver of informed consent.

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Footnotes

  • Handling editor Mary Dawood

  • Twitter @TimothyJang, @newgardc

  • Contributors SK, MDR, MC-S and EKC designed the study. The study was conceived through discussions involving SK, MDR, MC-S, SB, TBJ, HW, MH, GTL, JPd'E, CDN, EKC and MPL. SK analysed the data. SK and MDR worked on the first draft. All authors contributed to the writing, editing, reviewing and the preparation of the final draft. JAK conceived the RECOVER database, organised sites, obtained funding, collected data and edited and approved this article. SK is the guarantor.

  • Funding This project was funded by the Department of Emergency Medicine at the Indiana University School of Medicine.

  • Competing interests None declared.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.

Linked Articles