Diminished Health Returns of Employment During COVID-19 Pandemic

,

and SDoH indicators show their strongest effects on socially privileged groups who can easily mobilize their SES and SDoH indicators to tangible outcomes. 34,35 At the same time, SES and SDoH indicators have weaker effects for racialized and marginalized people. 34,35 This observation is summarized as marginalization-related diminished returns (MDRs) and minorities' diminished returns (MDRs). 36 As a result of these MDRs, economic and health inequalities extend from lower to middle-class America. 37,38 In line with these MDRs, some evidence suggests that some racial, economic, and health gaps may increase, rather than decrease, as SES increases. 39 These are in part because of structural racism, which can manifest in many shapes and forms including but not limited to segregation, 40 low education quality in urban areas, harsh school disciplinary actions, 41 unfair banking policies, 42 harsh policing, 43 and discrimination in the labor market, 44 all of which serves to block opportunities for minority populations across class lines. As such, establishing MDRs has become a tool to measure structural racism. 45,46 According to the MDRs literature, that has been exclusively conducted outside the COVID-19 pandemic, while high SES non-Latinx Whites show excellent health, high SES non-Latinx Blacks may remain at health risk. 47 This pattern is well established for parental education, 47 education, 48,49 and income, 50 however, there are only two studies on MDRs of employment on health. 51,52 While MDRs are shown for Latinx, 53 Asian, 50 and Native American 54 individuals, most of the literature is on Black vs. White individuals. 48 Therefore there is a need to test if the health effect of employment is similar for non-Latinx White in comparison to non-Latinx Black people, during the COVID-19 pandemic, while education and income are controlled. 47 In addition, although similar findings are shown for chronic disease, 55 obesity, use of cigarettes, 47 e-cigs, 49 Hookah, 56 and other substances, 57,58 and SRH, 59 all of these comparisons have been conducted in normal times without a macro event such as COVID-19 pandemic. They are also relevant to children, youth, adults, elders and various sources of marginalized groups based on race, 47 ethnicity, 60,61 sexual orientation, 62 nativity, 63,64 and even place, 65 suggesting that any marginalization in the society reduces the gains that are expected to follow SES and SDoH on health. As a result of these increased health risks, we observe higher than expected risk of asthma, 66 chronic obstructive pulmonary disease (COPD), 67 and heart disease 55 in high SES racialized adults.
There is a need for additional research on the MDRs of SES indicators, such as employment, with health outcomes such as SRH across diverse racial groups during the COVID-19 pandemic. Almost all past research is conducted in an era when pandemics do not limit living conditions. During the COVID-19 pandemic, SES indicators had an important role, and employment could impose or protect risk for individuals. As unemployment is higher for Blacks than Whites and Blacks typically have lower financial security, which could mean a higher reliance on the continuation of their job, it is important to test the employment-SRH link between White and Black adults during the COVID-19 pandemic.
We conducted this study to test the association between employment and SRH overall and by race. As employment is also confounded by education and income, we are interested to control for income and education. This will help us go beyond independent effects of employment and also test additive effects of employment, income, and education. We hypothesized an inverse association between employment and poor SRH, however, we expect this association to be stronger in non-Latinx Whites than non-Latinx Black individuals. In line with MDRs outside COVID-19 era, we expected employment, as a major economic resource, to have weaker health effects for Black people, as a historically marginalized group, during the COVID-19 pandemic.

Methods
This secondary data analysis applied a cross-sectional methodological design. Data came from the Health Information National Trends Survey (HINTS 2020) study Cycle 4 which was conducted between February 24, 2020, and June 15, 2020. Given the data's de-identified nature, our secondary analysis was exempt from a full ethics review.
The HINTS study participants were adults residing across US states in 2020. The sampling frame for Cycle 4 consisted of a database of addresses used by Marketing Systems Group (MSG) to provide random samples of addresses. Any nonvacant US residential address was subject to sampling. This included but not limited to those present on the MSG database, including post office (P.O.) boxes, throwbacks (i.e., street addresses for which mail is redirected by the United States Postal Service to a specified P.O. box), and seasonal addresses. Although a total number of 3865 individuals completed surveys, which resulted in a 37% response rate, for this analysis, we only included 1403 individuals including 1109 (79%) non-Latinx White and 294 (21%) non-Latinx Black participants who had complete data on employment, education, income, age, sex, marital status, and self-rated health (SRH) and were either non-Latinx White or non-Latinx Black. We only included those who were recruited after pandemic was announced by the World Health Organization (WHO) and the Centers for Disease Control and Prevention (CDC).
The HINTS 2020 used a multi-stage stratified random sampling. For first sample stage, the sampling frame of addresses was grouped into the following two explicit sampling strata: 1). Addresses in areas with high concentrations of minority population; and 2). Addresses in areas with low concentrations of minority population. The second sampling stage was selection of a participant from each selected household. Only up to one participant was selected from each target household, upon eligibility.
SRH. The dependent variable was poor subjective SRH, measured by the following conventional SRH item: "In general, would you say your health is…" Item responses included excellent, very good, good, fair, or poor. We considered the answer "poor" as poor SRH (score = 1) and excellent, very good, good, and fair as good SRH (score = 0). So, our outcome reflected poor not good health.
Race. Participants were asked if they were White, Black, or from other racial background. The question read as "Are you Black or African American?" All participants who positively answered to the last question were excluded from this analysis.

Independent Variables
Employment Educational attainment. The first independent variable was highest level of education at the individual level, measured by self-reported educational attainment question. We calculated this variable based on the highest level of education which was attained. The specific item was "What is the highest grade or level of schooling you completed?". This variable was a continuous variable with the following seven categories.
This variable was a continuous variable ranging from 1 to 9. Marital status. The individual disclosed family marital status, a dichotomous variable which was coded as married or non-married (reference category). The specific item read as "What is your marital status?" Gender. A dichotomous variable, gender was coded as male = 1 and female = 0 (reference category). Gender was self-reported.
Age. Participants reported their age. Age was a continuous variable measured in years. The question read as "What is your age?".
Using SPSS 21, we performed univariate, bivariate, and multivariable analysis. For univariate analysis, we reported the mean (SD) and frequency tables (%) for our variables overall and by racial group. We calculated Chi-square and t test to compare our study variables by racial group, for our bivariate analysis. For our multivariable analysis, logistic/ linear regression models were estimated for each independent variable (education or income). The first models did not include any interaction terms. These models only included main effects of race, employment, education, income, and covariates. After running our Models 1, Models 2 were performed that also had race by employment interaction term. This model included all previous terms (main effects) in addition to one race by employment interaction term. To test our modeling assumptions, we ruled out collinearity between study variables particularly education, income, employment, and race. The independent variable was employment, covariates included education, income, gender, age, and marital status. The moderator was race, as a proxy of racialization because we had controlled various SES indicators. Odds ratio (OR), regression coefficient, standard errors (SEs), and P values were reported. A P value of less than 0.05 was significant.

Results
Overall, 1403 individuals entered our analysis. This number included 1109 (79%) non-Latinx White and 294 (21%) non-Latinx Black participants. Table 1 reports descriptive data overall and by race. Participants varied in age from 18 to 100 years old. Table 2 provides a summary of binary logistic regression models without and with interaction term between employment and race. According to this table, based on Model 1, employment was inversely associated with poor SRH, meaning that adults who were employed reported better SRH. According to Model 2, however, the employment-SRH association varied by race, with the inverse association being weaker for non-Latinx Black than non-Latinx White individuals. Table 3 provides summary of logistic regressions by race. The results of previous Model 2 were confirmed, meaning a weaker protection of employment against poor SRH for non-Latinx Black than non-Latinx White individuals. While the protection was significant for non-Latinx Whites, it was not significant for non-Latinx Blacks.

Discussion
The aim of this study was to test overall and racial differences in the association between employment and SRH among American adults during the COVID-19 pandemic. Our first hypothesis was there would be an inverse association between employment and poor SRH, which was reflective of better health of employed than unemployed people. Our second hypothesis was that the protective effect of employment against poor health would be stronger for non-Latinx White than non-Latinx Black individuals. Both of our hypotheses were confirmed. Similar to education and income, employment is a major SDoH and SES indicator. 1,2 As shown by Mirowsky and Rossm 3-5 Marmot, 6-11 Link and Phelan, [12][13][14][15] and House and Lantz, [16][17][18] and others, 19 health effects of SES indicators hold across populations, outcomes, settings, and age groups. Their work has generated robust empirical evidence and rich theoretical argument on better health of individuals who are employed, educated, and have higher income. These SDoHs and SES indicators enhance a wide range of health, behavioral, and developmental outcomes through various mechanisms that including better environment, healthy options, healthy choices, low stress, and healthy development. [20][21][22][23][24][25][26] However, some SES indicators operate through more behavioral and some SES indicators may operate through developmental and contextual mechanisms. 27,28 For example, families with higher SES and SDoH resources show less substance use, 29,30 SRH, 59 and depression. 68 Over time, SES indicators and SDoHs are showing stronger health effects. 29 In line with MDRs, for non-Latinx Whites, poor health is concentrated for poor, unemployed, and less educated people. [31][32][33] This is not the case for Blacks for whom SES indicators are less salient, given diminished returns of SES. 59 For Blacks, health problems sustain across class lines, because SES and class show weaker health effects.
We found that in COVID-19 pandemic, an SES and SDoH indicator such as employment may have weaker effects for racialized and marginalized people, particularly non-Latinx Black people. 34,35 This observation is in line with MDRs  theory. 36 As a result of these MDRs, economic and health inequalities extend from lower to middle-class America. 37,38 In line with these MDRs, some evidence suggests that some racial, economic, and health gaps may increase rather than decrease as SES increases. 39 Due to structural racism, Jim Crow, social stratification, historic discrimination, and residential and job segregation, as well as labor market discrimination, non-Latinx Black individuals work in worse jobs than non-Latinx White people, which may reduce the health return of employment for Black communities. This is why some scholars have indicated that MDRs reflects structural racism. 45,46 Racism can manifest in many shapes and forms including, but not limited to, segregation, 40 low job and education quality in urban areas, 44,69-74 unfair banking policies, 42 harsh policing, 43 and discrimination in the labor market. 44 All of these processes may further block opportunities for minority populations across class lines (regardless of their employment and SES). This is why MDRs should be undone if we wish to undo racism. 36 In one study, employed non-Latinx people were protected against tobacco use, but employed Latinx people had high risk of smoking. 51 In another study, highly educated non-Latinx Black people had high occupational stress, while highly educated non-Latinx White people had low occupational stress. 75 Finally, in a study, employment had a larger effect on life expectancy of Black than White people. 76 These MDRs reflect unequal occupational opportunities of Whites and Blacks, regardless of SES indicators such as employment. 59 In the US, people's insurance status is closely tied to their employment. Thus insurance may be a factor in MDRs of employment for Black populations.
A recent piece of literature on MDRs has shown that, while high SES non-Latinx Whites show the least health problems, high SES Latinx and Black people report higher levels of poor health and risky behaviors. 47 This association is reported for parental education, 47 education, 48,49 and income. 50 Similarly, the same finding is shown for mental, 68 physical health, 67 and health behaviors such as traditional cigarette, 47 e-cig, 49 Hookah 56 and alcohol use, suggesting that these diminished returns are independent of health problems or risk behaviors. 77 They are also shown for youth, adults, and older adults, as well as various sources of marginalization, namely race, 47 sexual orientation, 62 and immigration status, 63,64 suggesting that any marginalization in the society reduces the health gains that are expected to follow SES and SDoH. To give a few examples, we observe higher than expected risk of asthma, 66 COPD, 67 and heart disease 55 in high SES racialized and racial minority adults.
There is a need for additional research on the effect of time, cohort, pandemics, and other political and macro factors on the associations between SES indicators and health outcomes across diverse racial groups. Most past research is conducted regardless of macro events such as pandemics or economic slowdowns. These macro events may have differential impacts on subpopulations, and there is a need to compare White and Black individuals across time frames. Thus, there is a need to compare diverse groups for the health returns of SES indicators across time intervals that may change human and economic behaviors. The COVID-19 pandemic, for example, resulted in a major pressure across minority populations. [78][79][80][81][82] While MDRs are also shown for Latinx, 53 Asian, 50 native American, 54 immigrant, 63,64,[83][84][85] and Black 48 individuals, almost all of this literature is on normal times. 48 So there is a need to test the effects of COVID-19 pandemics in changing the recognized patterns for diverse populations. Mechanisms of disparities may change based on macro data, and contributors of health disparities may vary across time. 47

Limitations
There are some limitations to the current study. The sample size was different across racial subgroups, thus the statistical power was non-identical across racial groups. The outcome was single item self-reported, which may reflect measurement bias by race. Experience and report of SRH may be influenced by race, culture, SES, and sex/gender. We excluded Latinx, Asian, and other marginalized groups. We also did not have data on type of job, years of experience, and pay per hour/ year, that could reflect labor market discrimination. All our variables were individual level, and we did not have access to distribution of jobs and occupational segregation in neighborhoods. Some strengths include large overall sample size, robust methodology, and random sample, and control of other SES indicators such as education and income.

Conclusion
To conclude, employment, as a SES indicator, shows diminished health returns for marginalized and racialized people (non-Latinx Black), which may reflect racism, social stratification, and historic discrimination in the US. This observation holds for the COVID-19 era, and addressing health inequalities during the pandemic requires addressing MDRs.

Authors' Contributions
Study design: AA, data collection: BN, conceptual design: SA, data analysis: SA, prepare draft: AA, revision: AA, BN, SA. All authors approve the final draft.

Conflict of Interest Disclosures
The authors declare that they have no conflict of interest.

What Is Already Known?
Employment protects against poor health. Individuals who are employed report better health. The association between employment and health is not universal.

What Does This Study Add?
There are racial differences in the link between employment and health. While employed White people are healthy, employed Black individuals report poor health. Social stratification, racism, and discrimination may reduce the health benefits of employment in Black communities.

Ethical Approval
It was exempt from a full IRB review because it was based on a fully deidentified publicly available data set.
Funding/Support SA, the corresponding author, is supported by the National Institutes of Health (NIH) grant 5S21MD000103.