{"subscriber":false,"subscribedOffers":{}}

Cookies Notification

This site uses cookies. By continuing to browse the site you are agreeing to our use of cookies. Find out more.
×

Research Article

COVID-19

COVID-19 And Racial/Ethnic Disparities In Health Risk, Employment, And Household Composition

Affiliations
  1. Thomas M. Selden ([email protected]) is director of the Division of Research and Modeling, Center for Financing, Access, and Cost Trends, Agency for Healthcare Research and Quality, in Rockville, Maryland.
  2. Terceira A. Berdahl is a social science analyst in the Division of Research and Modeling, Center for Financing, Access, and Cost Trends, Agency for Healthcare Research and Quality.
PUBLISHED:Free Accesshttps://doi.org/10.1377/hlthaff.2020.00897

Abstract

We used data from the Medical Expenditure Panel Survey to explore potential explanations for racial/ethnic disparities in coronavirus disease 2019 (COVID-19) hospitalizations and mortality. Black adults in every age group were more likely than White adults to have health risks associated with severe COVID-19 illness. However, Whites were older, on average, than Blacks. Thus, when all factors were considered, Whites tended to be at higher overall risk compared with Blacks, with Asians and Hispanics having much lower overall levels of risk compared with either Whites or Blacks. We explored additional explanations for COVID-19 disparities—namely, differences in job characteristics and how they interact with household composition. Blacks at high risk for severe illness were 1.6 times as likely as Whites to live in households containing health-sector workers. Among Hispanic adults at high risk for severe illness, 64.5 percent lived in households with at least one worker who was unable to work from home, versus 56.5 percent among Black adults and only 46.6 percent among White adults.

TOPICS

In many areas of the US, non-Hispanic Blacks and Hispanics are more than twice as likely as non-Hispanic Whites to die from coronavirus disease 2019 (COVID-19).15 The Centers for Disease Control and Prevention (CDC) found that age-adjusted COVID-19 hospitalization rates for Blacks and Hispanics as of May 30, 2020, were 4.5 and 3.5 times those of Whites, respectively.6 The magnitude of these disparities has focused renewed attention on the consequences of long-standing structural inequality along racial/ethnic lines with respect to a wide range of outcomes including income, health, health care, employment, and living circumstances.715 In this article we use data from the Medical Expenditure Panel Survey (MEPS) to recast and build on pre-COVID-19 disparities research, offering insights into a number of hypotheses that have been proposed with regard to the causes of COVID-19 disparities.

One common hypothesis is that disparities in COVID-19 outcomes arise from preexisting differences in underlying health conditions that increase the severity of COVID-19 illness for Blacks and Hispanics, conditional on exposure to the virus. To explore this hypothesis, we examined racial/ethnic differences in health risks that the CDC associates with severe COVID-19 illness. Because many reported statistics regarding COVID-19 disparities have not been age adjusted, we examined health risks both without age adjustment and by age group, as is more conventional in the disparities literature. Estimating risks in both ways yields important insights, because age itself is a key COVID-19 risk factor and because average age varies considerably across racial/ethnic groups.

We also used MEPS data to examine the hypothesis that COVID-19 disparities stem from racial/ethnic differences in employment-related risk for infection. Prior research shows that compared to White workers, Black and Hispanic workers are more likely to work in jobs that have a higher risk for exposure to infectious disease9,16 and that have, in many cases, been deemed essential during the pandemic. Blacks and Hispanics are also less likely to be able to work from home.17 Using MEPS data, we were able not only to replicate these findings but also to explore how disparities in employment intersect with differences in health risk, both among workers and among those adults with whom workers live.

Study Data And Methods

MEPS is a nationally representative survey of the civilian noninstitutionalized population, sponsored by the Agency for Healthcare Research and Quality. We pooled data from survey years 2014–17, yielding 100,064 person-year observations on adults ages eighteen and older (see online appendix exhibit 1).18 Estimates were weighted to be nationally representative of the civilian noninstitutionalized population and its racial and ethnic subgroups. Standard errors and statistical tests were corrected for the complex design of MEPS. Race and ethnicity were defined for respondents whose single reported race was White, Black, or Asian and for respondents of Hispanic ethnicity (any race). (In the article, White, Black, and Asian descriptions are assumed to be non-Hispanic.) The appendix presents additional technical details, additional results (as noted below), and point estimates and survey-adjusted standard errors for all exhibits, and appendix exhibit 1 provides selected background racial/ethnic disparity estimates on age, education, poverty, insurance, and household size.18

Health Measures

Our analysis followed the CDC’s April 2, 2020, guidance regarding health factors placing people at high risk for severe COVID-19 illness.19 We focused on treated conditions to reduce, in principle, the chance of attributing elevated COVID-19 risk on the basis of conditions that do not themselves merit treatment. In practice, our results were not sensitive to this decision. The CDC health risks we included were extreme obesity (body mass index higher than 40 kg/m2), current smoker, age sixty-five or older, diabetes (if treated with oral or injectable medication), and any of the following conditions if linked to current-year treatment: asthma, emphysema or other chronic obstructive pulmonary disease, cancer (other than nonmelanoma skin cancers), or coronary heart disease. We were only partially able to include kidney disease because of the difficulty of identifying cases stemming from a primary diagnosis of diabetes or high blood pressure using household-reported data. We did not implement the portions of the CDC guidance concerned with long-term use of immunosuppressant medications (in part because some immunosuppressants are being considered as COVID-19 treatments) or immunosuppressed conditions (other than cancers with ongoing treatment). We also did not include liver disease because of our inability to distinguish between chronic and acute cases. Because high blood pressure can be an important underlying condition affecting the severity of COVID-19 illness, we present estimates of its treated prevalence across race and ethnicity, and we included it in our counts of multiple conditions among those with high risk. However, following CDC guidance at the time of our analysis, we did not include high blood pressure in our summary CDC measure of people at high risk for severe COVID-19 illness.

Job Characteristics

We augmented MEPS data with information on ability to work from home from the 2017–18 American Time Use Survey,20 sponsored by the Bureau of Labor Statistics. We used weighted sequential cold-deck imputation by detailed industry and occupation, as well as race/ethnicity, age, sex, and education. To account for the rapid changes COVID-19 prompted in how America works, we assumed that all education workers would be able to work from home, whereas workers in hospitals, public safety, and selected other industries would be unable to do so. To identify essential workers, we applied the Department of Homeland Security’s Cybersecurity and Infrastructure Security Agency guidelines from March 28, 2020, regarding essential workers.21

Limitations

This study was subject to several limitations. Our analysis examined only the community noninstitutionalized population, thereby excluding hard-hit populations in nursing homes, long-term care facilities, and correctional facilities. A recent estimate is that 42 percent of all COVID-19 fatalities have occurred among residents of (and in some cases workers at) nursing homes and long-term care facilities.22

We measured health risks using self-reported data on conditions and treatments. However, if conditions among Blacks and Hispanics were less likely to be diagnosed or less likely to be treated if diagnosed, then actual disparities in condition prevalence might have been larger than we observed.14 A recent analysis of data from the National Health and Nutrition Examination Survey found a 2.0-percentage-point difference in undiagnosed diabetes between Blacks and Whites (the corresponding White–Hispanic difference was 1.1 percentage points).8 Black–White differences were smaller for the other conditions studied. These differences are small compared with the magnitude of the disparities being found in COVID-19 outcomes. Moreover, including body mass index over 40 kg/m2 and age sixty-five or older as risk factors might have helped us capture some of the cases with undiagnosed conditions (because of the correlation of body mass index and age with a number of other conditions on the CDC list).

Another limitation is that our data reflect employment before the onset of COVID-19. They therefore provide no insights into whether there have been racial/ethnic dimensions to the COVID-19 employment changes. For instance, there may have been racial/ethnic disparities in the ability to leave a job (perhaps reflecting other household income sources and wealth);23 the extent to which employers laid off certain groups of workers; or the extent to which workers who were laid off found alternative employment, perhaps in sectors such as home delivery that have rapidly grown during the pandemic.

Study Results

Health Risk

Within each age group, MEPS results mirror well-known findings on health disparities, with Black and Hispanic adults being more likely than White adults to suffer from conditions that the CDC associates with higher risk for severe COVID-19 illness (for results by age group, see appendix exhibit 2b).18 However, controlling for age when examining health differences creates a mismatch when the goal is to understand COVID-19 disparities that are, for the most part, not age adjusted. White adults were older, on average, than minority adults, with an average age of 50.0 years, versus 45.0, 44.7, and 41.5 years, respectively, for Black, Asian, and Hispanic adults (appendix exhibit 1).18 Given the importance of age, both as a correlate of health risks and as a COVID-19 risk factor in its own right, exhibit 1 presents unadjusted or population average estimates.

Exhibit 1 Prevalence among US adults of selected risk factors and an overall Centers for Disease Control and Prevention (CDC) indicator for being at high risk for severe COVID-2019 illness, by race and ethnicity, 2014–17

Exhibit 1
SOURCE Authors’ calculations using data from the Medical Expenditure Panel Survey (MEPS), 2014–17. NOTES Health risks shown in exhibit are body mass index (BMI) higher than 40 kg/m2, based on self-reported height and weight, with missing data filled in from the same person in other years of MEPS or linked National Health Interview Survey data; current smoker; age sixty-five or older; diabetes (if treated with oral or injectable medication, either from the MEPS Diabetes Care Supplement or, for supplement nonresponders, from MEPS data on prescription drug use); high blood pressure (if treated with prescription medicine); and the following conditions if linked to current-year treatment: asthma, cancer (other than nonmelanoma skin cancers), or coronary heart disease. Health risks are not mutually exclusive. The CDC high risk indicator is based on having at least one of these factors, other than high blood pressure, or having either emphysema or other chronic obstructive pulmonary disease or kidney disease (not shown). Asterisks indicate that the estimate is statistically significantly different from the non-Hispanic White estimate at the indicated levels. ***p<0.01

The pre-COVID-19 health disparities in exhibit 1 seem too small to explain the twofold or larger disparities being reported in COVID-19 mortality. Indeed, Whites were more likely than Blacks, and far more likely than Asians or Hispanics, to be identified as high risk by virtue of having at least one CDC risk factor.

Although the CDC guidelines identify patients as high risk on the basis of any of the listed conditions, patients with multiple conditions are likely to be at even greater risk. Exhibit 2 presents evidence on the average number of risk factors and the distribution of these risk factors by race and ethnicity. The share of Whites having at least three conditions was 25.2 percent, versus 24.4 percent for Blacks (difference not statistically significant) and less than 20 percent for Asians and Hispanics (all estimates include high blood pressure in their risk factor counts).

Exhibit 2 Number of risk factors (including high blood pressure) among US adults at high risk of severe COVID-19 illness, by race and ethnicity, 2014–17

Average no. of risk factorsShare of high-risk adults with 1 risk factor (%)Share of high-risk adults with 2 risk factors (%)Share of high-risk adults with 3 or more risk factors (%)
All adults1.9045.430.524.2
Non-Hispanic White1.9344.330.625.2
Non-Hispanic Black1.9243.731.924.4
Non-Hispanic Asian1.78***47.133.3*19.7***
Hispanic1.72***53.1***28.1***18.8***

SOURCE Authors’ calculations using data from the Medical Expenditure Panel Survey, 2014–17. NOTES Definitions of risk factors and high risk are in the notes to exhibit 1. Asterisks indicate that the estimate is significantly different from the non-Hispanic White estimate at the indicated levels.

*p<0.10

***p<0.01

Not only were the magnitudes, and in some cases the directions, of the risk-factor differences in exhibit 1 at odds with the twofold and higher disparities being reported in COVID-19 fatalities, but also we observed a pattern across race/ethnicity and by sex that seemed at odds with what little is known about COVID-19 outcomes on these dimensions. Exhibit 3 shows that the difference between Black and White women was larger than the difference between Black and White men for obesity, diabetes, asthma, and high blood pressure. Being a current smoker was the only risk factor for which the Black–White difference was larger for men than for women. If these differences in health risk disparities by sex were important drivers of COVID-19 outcomes, we would expect Black–White COVID-19 outcome disparities to also be larger among women than among men. In contrast, appendix exhibit 4 presents estimates from the CDC COVID-19-Associated Hospitalization Surveillance Network database showing no corresponding differences by sex in racial/ethnic disparities in cumulative hospitalizations as of May 30, 2020.6,18

Exhibit 3 Selected COVID-19 risk factors among US adults, by race/ethnicity and sex, 2014–17

BMI >40 kg/m2 (%)Diabetes (%)Asthma (%)Current smoker (%)Ever diagnosed with high blood pressure (%)Age 65 or older (%)Meets CDC definition of high risk for severe COVID-19 illness (%)
Men
Non-Hispanic White4.010.03.615.427.122.946.7
Non-Hispanic Black4.6*11.6**3.0*19.3***26.013.8***42.3***
Non-Hispanic Asian0.5***10.52.1***9.9***17.9***14.6***31.1***
Hispanic3.78.8**2.3***10.1***13.9***9.1***28.4***
Women
Non-Hispanic White5.98.56.312.625.025.948.7
Non-Hispanic Black12.4***12.7***7.4**12.532.1***16.7***46.7*
Non-Hispanic Asian0.5***7.2*2.7***2.2***15.9***15.9***23.5***
Hispanic5.810.2***4.7***5.6***15.2***11.9***30.7***

SOURCE Authors’ calculations using data from the Medical Expenditure Panel Survey, 2014–17. NOTES Definitions of risk factors and high risk are in the notes to exhibit 1. Asterisks indicate that the estimate is significantly different from the (same sex) non-Hispanic White estimate at the indicated levels. BMI is body mass index. CDC is Centers for Disease Control and Prevention.

*p<0.10

**p<0.05

***p<0.01

Health Risk Sensitivity Analysis

The results on health risks presented in exhibit 1 are national averages, whereas the impact of COVID-19 has been unequally distributed across the country. Appendix exhibit 5 presents results stratified by region for Metropolitan Statistical Areas (MSAs) and nationwide for non-MSAs.18 In all cases, we found similar patterns, with minority health status being either only modestly worse than among Whites or, in some cases, better than among Whites, according to the CDC measure.

Appendix exhibit 6 presents additional sensitivity results.18 The top row presents our main estimates at the national level for adults facing high risk overall and by race/ethnicity. The second row excludes age as a risk factor. This lowered the overall prevalence from 43.2 percent to 33.3 percent. It also had a disproportionate impact on the percentage of Whites identified as high risk, changing the sign of the Black–White difference. However, the difference between Blacks and Whites remained small relative to the large differences being reported in COVID-19 mortality. The third row of the exhibit responds to studies questioning the inclusion of current smoking24 and asthma25 as risk factors for severe COVID-19 illness. Excluding smoking and asthma as risk factors also lowered the overall prevalence (to 31.8 percent) and modestly widened the Black–White gap. In the fourth row, factoring in untreated conditions had little effect on either the overall prevalence of high risk or its distribution across race and ethnicity. The same was also true when we relied on MEPS priority conditions. In all of the measures, we observed substantial differences in risk between Whites and Blacks versus Hispanics and Asians, with Whites and Blacks having greater prevalence of high risk.

Employment Risks

Exhibit 4 presents the distribution of workers by race and ethnicity across a number of COVID-19-relevant job characteristics. Essential businesses were those most likely to stay open for business despite public policy on social distancing and concerns about infection. Minorities were only slightly more likely than Whites to work in essential jobs. The percentage of White workers who worked in essential jobs was 75.7 percent, versus 78.3 percent, 80.6 percent, and 78.0 percent among Black, Hispanic, and Asian workers, respectively (see appendix exhibit 7).18 Blacks were substantially more likely than Whites to work in the health care sector (16.3 percent versus 10.4 percent).16 Blacks were also overrepresented in public safety, whereas Hispanics were heavily overrepresented in the food sector.9 Among White workers, 22.8 percent were essential workers who were able to work from home versus only 13.3 percent of Black workers and 12.5 percent of Hispanic workers. These differences were smaller among nonessential workers.

Exhibit 4 Job characteristics among US workers, by race and ethnicity, 2014–17

Exhibit 4
SOURCE Authors’ calculations using data from the Medical Expenditure Panel Survey (MEPS), 2014–17. NOTES MEPS data were augmented with data on ability to work at home from the American Time Use Survey, 2017–18. Health care workers are those employed in hospitals, outpatient clinics, physician offices, nursing homes, and other residential treatment facilities; home health care workers; and ambulance drivers. Public safety workers include police officers, firefighters, corrections workers, public transportation employees, postal workers, and workers at funeral parlors and crematoriums. Asterisks indicate that the estimate is significantly different from the non-Hispanic White estimate at the indicated levels. **p<0.05 ***p<0.01

These results for job characteristics suggest the possibility of disparities in employment-related infection risk across race and ethnicity. However, differences in these patterns by sex complicate this interpretation. Appendix exhibit 8 stratifies the estimates in exhibit 4 by sex, showing that much of the racial/ethnic difference in health-sector employment occurred among women.18 Among Black women, 24.6 percent worked in health care versus only 17.2 percent of White women. The corresponding shares among men were 7.1 percent and 4.4 percent, respectively. In contrast, disparities in the percentage of workers in essential jobs who can work from home were larger among men: 25.6 percent among White men and 12.9 percent among Black men, versus 19.7 percent among White women and 13.5 percent among Black women.

Household-Level Analysis

In the US, racial/ethnic minorities often reside in households with more members, relative to Whites. This pattern has implications for social distancing because workers’ risks for COVID-19 exposure can become risks shared by all household members with whom they live. In MEPS, the average household size for Whites was 2.8 people, versus 3.1 for Blacks and 3.8 for Hispanics (appendix exhibit 1).18 To show how health and employment differences intersect with household composition differences, we examined the distribution of adults at high risk for severe COVID-19 illness by race/ethnicity across households classified according to the employment status and job type of all household members. In exhibit 5, 15.1 percent of high-risk Blacks lived in households with at least one worker in the health sector versus 9.3 percent of Whites. If we add together the percentages in exhibit 5 for high-risk adults living with at least one worker in health, public safety, public utility, food, or other jobs lacking the ability to work from home, we obtain the percentage of high-risk adults living with a person who cannot work from home. We found that 56.5 percent of high-risk Blacks lived in households with at least one worker who was unable to work from home versus 46.6 percent of high-risk Whites. Among Hispanics, the frequency of living with at least one worker in the health sector was lower than among Blacks; however, the overall frequency of living with at least one worker who cannot work from home was even higher (64.5 percent).

Exhibit 5 Distribution of US adults at high risk for severe COVID-19 illness, by job characteristics of workers in household and by race and ethnicity, 2014–17

Exhibit 5
SOURCE Authors' calculations using data from the Medical Expenditure Panel Survey, 2014–17. NOTES For the definition of high risk, see notes to exhibit 1. Descriptions of work categories are in the notes to exhibit 4. To be included in the count, a household had to include at least one person employed in the category listed. Asterisks indicate that the estimate is significantly different from the non-Hispanic White estimate at the indicated levels. **p<0.05 ***p<0.01

Discussion

Disparities in COVID-19 outcomes likely stem from structural racism on many levels.

Disparities in COVID-19 outcomes likely stem from structural racism on many levels, including income, education, health insurance, access to medical care, access to food, health status, job characteristics, living conditions, and more. Simply listing these potential factors, however, falls short of providing a clear explanation for the twofold and larger disparities in adverse COVID-19 outcomes that are being reported in the US.

Two broad groups of explanations have been offered for COVID-19 disparities: one focusing on differences in risk for infection (for example, through work exposure, household transmission, or community contact) and the other focusing more on differences in illness severity conditional on infection (for example, as a result of preexisting health risks). Data on random testing could help clarify the relative importance of differences in infection rates versus differences in illness severity conditional on infection; however, to our knowledge, such data do not exist. This information gap motivated us to analyze prepandemic data from MEPS, recasting and building on well-known results from the disparities literature to provide insights of relevance to understanding the racial/ethnic dimensions of COVID-19.

Much attention has been given to the hypothesis that racial/ethnic differences in the health factors that the CDC has associated with severe COVID-19 illness cause minorities to have worse outcomes conditional on infection. As in prior research, we showed that if age is held constant, then Blacks and, to a lesser extent, Hispanics tended to have a higher prevalence of disease and other health risk factors than Whites. These differences, however, seem too small to explain the disparities of unprecedented magnitude being reported for COVID-19 outcomes. Moreover, Whites were older, on average, than other racial/ethnic groups, and this age difference tended to narrow or even reverse health disparities when racial/ethnic groups were compared as a whole (without age adjustment). For this reason, and because the CDC has identified age itself as a COVID-19 risk factor, the frequency of being at high risk for severe COVID-19 illness was actually higher among Whites than among Blacks, Hispanics, or Asians. Further complicating the health differences hypothesis is that Black–White health differences, to the extent that they exist, are generally larger among women than among men, a pattern we do not see in the available evidence on COVID-19 outcomes. We conclude that disparities in the CDC risk factors appear unlikely by themselves to explain high COVID-19 morbidity and mortality among minorities.

The second strand of our research focused on dimensions of employment and household composition that might contribute to differences in rates of infection. We observed large racial/ethnic differences in job characteristics, with Blacks being substantially more likely than Whites to work in health, public safety, and public utility jobs. Within racial/ethnic groups, more Black women were employed in the health sector, a disparity that is not mirrored in COVID-19 hospitalization rates by race/ethnicity and sex. Hispanics were much more likely than Whites to work in food-related jobs, and as in the prior literature, both Blacks and Hispanics were less likely than Whites to be able to work from home.17 Especially when we factor in the much larger Black and Hispanic household sizes, a picture emerges in which employment may be an important pathway for infection of minority workers and their household members.

More research is needed to fully understand racial/ethnic disparities in COVID-19 infections and outcomes. None of the differences we examined in this article offers a full explanation for the large disparities in COVID-19 hospitalizations and mortality that have been reported. One area for future research would be to move beyond the health measures highlighted in the CDC guidance (and measured in MEPS) to examine the role in COVID-19 outcomes played by other health-related factors. For example, it may be important to look for racial/ethnic differences in the severity of preexisting conditions that may stem from structural inequalities in poverty, insurance, and access to care. Our analysis of multiple conditions revealed no evidence of severity differences on that dimension, but the role of unmeasured condition severity and the extent to which conditions are well controlled remain important open questions. Another health-related area for research concerns the psychosocial stressors that minorities can face as a result of racism and discrimination (allostatic load) and the associated impacts on cytokine response that may play important roles connecting the lived experiences of minorities to their physiological responses to COVID-19 infection and treatment.10,11,15,26,27 Yet another health-related topic concerns the quality of care received by White versus minority patients as they sought care for COVID-19, if only to understand how care was affected by geographic variation in the concentration of cases.

In addition to further research on the determinants of illness severity conditional on infection, our analysis suggests the importance of additional research on factors affecting exposure. Further insights can be gained from probing detailed job characteristics, such as occupational differences within health care. More can also be learned by exploring the extent to which spread through community contacts magnified the employment and household differences we observed in the MEPS data.

Conclusion

Given the paucity of racial/ethnic data on COVID-19, we examined historical patterns in health risk, job characteristics, and household composition in search of potential explanations for the large disparities being reported in COVID-19 outcomes. Our analysis highlights the need for improved real-time data, ranging from systematic testing and surveillance to improved reporting of race and ethnicity for hospitalizations and deaths (both in hospitals and in the community). That being said, given that health differences were incompatible in terms of size (and sometimes direction) with the sheer magnitude of the racial/ethnic differences in outcomes being reported, we believe that COVID-19 disparities will ultimately be shown to stem from disparities in exposure, such as the dimensions of employment and household transmission we examined and the dimensions of community contact that were outside the scope of this study.

ACKNOWLEDGMENTS

The authors appreciate the assistance of Francis Chesley, Joel Cohen, Doris Lefkowitz, Thomas Hegland, Julie Hudson, James Kirby, Julia McQuillan, David Meyers, Edward Miller, and Rosalie Torres Stone. All remaining errors belong to the authors. The views in this paper are those of the authors, and no official endorsement by the Department of Health and Human Services or the Agency for Healthcare Research is intended or should be inferred. An unedited version of this article was published online July 14, 2020, as a Fast Track Ahead Of Print article. That version is available in the online appendix.

NOTES

   
Loading Comments...