Racial and ethnic differentials in COVID-19-related job exposures by occupational standing in the US

Noreen Goldman; Anne R Pebley; Keunbok Lee; Theresa Andrasfay; Boriana Pratt

doi:10.1371/journal.pone.0256085

. 2021 Sep 1;16(9):e0256085. doi: 10.1371/journal.pone.0256085

Racial and ethnic differentials in COVID-19-related job exposures by occupational standing in the US

Noreen Goldman ^1,^*, Anne R Pebley ², Keunbok Lee ³, Theresa Andrasfay ⁴, Boriana Pratt ⁵

Editor: Marlene Camacho-Rivera⁶

PMCID: PMC8409606 PMID: 34469440

Abstract

Researchers and journalists have argued that work-related factors may be partly responsible for disproportionate COVID-19 infection and death rates among vulnerable groups. We evaluate these issues by describing racial and ethnic differences in the likelihood of work-related exposure to COVID-19. We extend previous studies by considering 12 racial and ethnic groups and five types of potential occupational exposure to the virus: exposure to infection, physical proximity to others, face-to-face discussions, interactions with external customers and the public, and working indoors. Most importantly, we stratify our results by occupational standing, defined as the proportion of workers within each occupation with at least some college education. This measure serves as a proxy for whether workplaces and workers employ COVID-19-related risk reduction strategies. We use the 2018 American Community Survey to identify recent workers by occupation, and link 409 occupations to information on work context from the Occupational Information Network to identify potential COVID-related risk factors. We then examine the racial/ethnic distribution of all frontline workers and frontline workers at highest potential risk of COVID-19, by occupational standing and by sex. The results indicate that, contrary to expectation, White frontline workers are often overrepresented in high-risk jobs while Black and Latino frontline workers are generally underrepresented in these jobs. However, disaggregation of the results by occupational standing shows that, in contrast to Whites and several Asian groups, Latino and Black frontline workers are overrepresented in lower standing occupations overall and in lower standing occupations associated with high risk, and thus may be less likely to have adequate COVID-19 protections. Our findings suggest that greater work exposures likely contribute to a higher prevalence of COVID-19 among Latino and Black adults and underscore the need for measures to reduce potential exposure for workers in low standing occupations and for the development of programs outside the workplace.

Introduction

In the United States, Black, Latino, and Native American adults have experienced substantially higher rates of COVID-19 infection and mortality during 2020 than Whites and Asians [1–7]. Researchers and journalists argue that the differences are due, at least in part, to two work-related factors: (1) Black, Latino, and Native American workers are more likely to hold jobs that have to be done at their workplace rather than remotely, and (2) Latino and Black workers face greater risks of exposure to COVID-19 in their jobs than others [3, 8–18]. This argument is consistent with the long history and contemporary effects of structural racism on occupational segregation in the US [19–22]. In this paper, we investigate differences in the likelihood of work-related exposure to COVID-19 by race and ethnicity. Our goal is descriptive: we provide insights into the pandemic by presenting the size and scope of these racial and ethnic disparities among workers rather than by estimating causal models of their determinants.

It is difficult to quantify the importance of occupational vs. other exposures to the coronavirus. Most studies have used two approaches to argue that occupational risks to COVID-19 are higher for marginalized racial and ethnic groups than for Whites. We use a combination of these approaches. The most common strategy has been to look at racial/ethnic differences in recent (generally pre-pandemic) employment in industries or occupations that are considered essential or frontline during the pandemic [8, 23, 24]. As we discuss later, these earlier studies have been based on various definitions of essential and frontline. A second approach, sometimes used in conjunction with the first, has been to estimate relative risk of different groups during the pandemic from survey data about whether jobs entail high exposure to disease or infection and/or require close proximity to other people [9]. Based on the recent availability of death records for 2020 in some locations, several studies have employed an additional strategy. They have used occupational codes from death certificates to directly identify occupations with exceptionally high mortality during the pandemic period and examine differential mortality by race and ethnicity [18, 25].

We extend this work in several ways. First, we consider 12 racial/ethnic groups. Previous studies on occupations and COVID-19 have largely been limited to four broad racial/ethnic categories (White, Black, Latino and Asian), although workers in other racial/ethnic groups or in Asian and Latino subgroups may be disproportionately exposed to COVID-19 [6]. Second, we examine five distinct job characteristics that could expose individuals to a high risk of contracting COVID-19: exposure to disease/infection; proximity to others; face-to-face discussions; interactions with external customers or the public; and working mostly indoors. We also present separate results by sex because of vastly different occupational profiles for men and women.

Third, we examine potential risk of exposure separately by occupational standing (OS), defined here as the proportion of workers in each occupation with at least some college education. We use OS as a proxy for workers’ access to COVID-19 mitigation measures in the workplace. As health inequality research shows, those with higher socioeconomic status (e.g., OS, education or income) generally have access to a wider array of resources, including power and influence, to protect health than others do [26]. In the case of potential workplace exposure to COVID-19, higher OS workers are more likely: (a) to work for employers who voluntarily practice risk mitigation and provide PPE and other tools for workers to do so, (b) if necessary, to demand risk reduction measures and to have the bargaining power to obtain them, and (c) to understand (or learn about) COVID-19 transmission routes and to comply with risk reduction strategies or implement their own [26–28]. Occupational standing is closely linked to having greater control over working conditions [29–31]. For example, despite the high risks often faced by physicians, nurses and other health personnel at the start of the pandemic when PPE was not readily available, these higher OS workers are generally more likely to work in environments that take disease transmission risk mitigation seriously and they have more power and knowledge to insist on effective risk mitigation when it is not available. Employers of lower OS workers in meat processing factories, farms, and retail stores, for example, are less likely to provide risk mitigation and workers have less power and knowledge to demand appropriate measures. Perhaps it is no surprise that lower OS workers appear to face especially high exposure to COVID-19 in the workplace [17, 27].

Examining results by occupational standing leads to a clearer picture of racial/ethnic differentials in potential exposure to COVID-19 transmission. Our results best reflect the situation at the end of 2020 when this paper was originally submitted. Although the situation has changed with widespread vaccination, lower OS and Black and Latino populations and those who are socioeconomically disadvantaged are less likely to be fully vaccinated than others [32], so that the picture we describe remains, unfortunately, relevant.

Data

We use data from two sources. Information on employment status, type of current/most recent job or business during the past five years, years of education, sex and race/ethnicity come from the 2018 American Community Survey (ACS), which was the most recent large, nationally representative survey fielded prior to the pandemic and available at the time of this research. Access is provided by the Integrated Public Use Microdata Series (IPUMS) [33]. Variables related to work context are drawn from the Occupational Information Network (O*NET), a database of work characteristics by detailed occupation collected by the US Department of Labor [34].

O*NET obtains information about job characteristics–such as work-related activities, work environment, and skills required to do the job–for almost 1000 detailed occupations. We use data from O*NET version 24.3, which contains information collected through 2019 and was released in May 2020 [34]. We identified six variables from O*NET surveys of incumbent workers that capture hazardous job characteristics that may expose workers to SARS-CoV-2 [17, 27]–which appears to be transmitted primarily through aerosols and respiratory droplets when an infected person coughs, sneezes, talks or breathes near others [35]. The six workplace variables reflect five types of risk: one variable for each of exposure to infection, physical proximity to others, face-to-face discussions, and interactions with external customers and the public, and two variables that reflect working indoors. Note that, like other studies, we do not have a direct indicator of actual exposure to the coronavirus, nor can we observe COVID-19 infections among workers.

Approval for this research was obtained from the Institutional Review Boards at the University of California Los Angeles and Princeton University.

Variables

Race/ethnicity

Individuals were classified according to the following 12 race/ethnicity categories as reported by the head of the household in the ACS: White, Black, Latino, Chinese, Filipino, Vietnamese, Korean, other Asian, Native American (American Indian/Alaskan Native), Pacific Islander (Native Hawaiian/Pacific Islanders), mixed race, and other race/ethnicity. The three largest groups in the other Asian category are Indian, Japanese and Pakistani. In our analysis, all racial/ethnic categories except Latino exclude those who reported themselves as Latino/Hispanic in the ACS question on ethnicity.

Occupational standing

Neither our data nor any other nationally representative data that we are aware of contain information on the use of COVID-19 mitigation measures. In order to rank occupations, we use a well-established sociological indicator of occupational standing, known as “occupational education” [36] as a proxy for access to workplace risk mitigation measures. Occupational standing is measured at the occupation, not the individual, level. For each occupation, we calculate occupational education from the ACS as the percent of all workers in the occupation who completed at least one year of college education (note that this variable classifies individuals by the educational attainment of all workers in their occupation, not their own educational attainment). We use this measure of occupational standing or “OS” for simplicity, to divide the 409 occupations reported by ACS respondents in our analysis into quartiles, i.e., each group comprising about one-quarter of the occupations (not one-quarter of the workers). We chose to use occupational education rather than occupational income–another well-established indicator of occupational ranking–both because educational attainment is typically reported with greater accuracy than income or wages and because among those with a valid occupational code in the ACS data file provided by IPUMS, educational attainment information is complete while information on income is missing for one-fifth of individuals. The correlation between these two measures of OS across the occupations in the dataset is high (0.72), indicating that occupational income and occupational education would produce similar rankings of occupational standing (OS).

Frontline status

During the pandemic, many workers have been sheltered from occupational exposure to infection by working remotely, e.g., from home. To distinguish between workers who are more vs. less able to work remotely, studies have used varying criteria to define “essential” and/or “frontline occupations.” For example, some studies have used guidelines issued by the Department of Homeland Security (DHS) Cybersecurity and Infrastructure Security Agency (CISA) to define essential occupations or industries [9, 23, 24, 37], while others have defined a set of occupational categories without providing specific selection criteria [38]. In this analysis, we include only occupations classified as “frontline” based on the definition offered by Dingel and Neiman [39] and used by Blau et al. [24]: occupations in which one-third or fewer workers can feasibly work from home, ascertained from responses to 15 questions in O*NET. A total of 249 out of 409 occupations in our analysis are considered frontline according to these criteria.

We do not consider whether occupations are classified as essential or subject to lockdown because there has been enormous variation across states, localities, and time period in definitions and their application during the pandemic. For example, janitors, maids, bus drivers, retail sales workers, and personal care workers would not be classified as essential according to industry guidelines issued by CISA [24], yet many individuals in these occupations have likely been working away from home much or most of the time since March, 2020.

Defining high risk occupations

The five types of risk we consider from O*NET data–exposure to infection, physical proximity to others, face-to-face discussions, interactions with external customers and the public, and working indoors–are based on the following O*NET questions (in their original wording):

How often does your current job require that you be exposed to diseases or infection?
How physically close to other people are you when you perform your current job?
How often does your current job require face-to-face discussions with individuals and within teams?
In your current job, how important are interactions that require you to deal with external customers (as in retail sales) or the public in general (as in police work)?
How often does your current job require you to work outdoors, exposed to all weather conditions? [And:] How often does your current job require you to work outdoors, under cover (like in an open shed)?

Each of these questions has five possible categorical responses; the responses reflect frequency of exposure for questions 1,3, and 5; degrees of closeness for question 2, and importance of interactions for question 4.

For each of the first four risk indicators, we define high risk occupations as those for which the mean response in the O*NET data falls in the highest quartile (25%) of the full set of 409 occupations in the analysis. For the fifth risk indicator we use the two questions shown above reflecting the frequency with which employees work outdoors; the responses for these two questions are highly correlated. Because indoor work is higher risk than outdoor work ceteris paribus, we take the maximum of the O*NET values for these two variables and define high risk as the quartile of occupations with the lowest values, thereby identifying jobs with the lowest frequency of working outdoors as high risk.

Analytic strategy

In order to match occupations reported in the ACS with characteristics in O*NET, we converted the occupation code (Standard Occupational Classification or SOC) in O*NET into the four-digit 2010 census occupational code recorded in the 2018 ACS using a crosswalk provided at https://www.bls.gov/emp/documentation/crosswalks.htm. In cases where one census occupational code corresponded to multiple O*NET occupations, we took unweighted averages of the characteristics of the O*NET occupations and assigned these to the census occupational code. We linked occupations between the ACS and O*NET for all employed persons who reported holding a job in the ACS with the exception of those in the military (O*NET data were not available for military occupations, which were held by about 6000 individuals in the ACS); this linkage yielded about 1.9 million individuals in the ACS in 409 occupations. Approximately 88 percent of these individuals reported about the job they held during the past year with the remaining respondents reporting about the job held between one and five years prior to the survey.

It is important to note that unemployment and labor force participation rates vary considerably by race, ethnicity, and sex. For example, in the third quarter of 2020, the unemployment rate for Black men 16+ years old was 13.8 percent compared to 7.4 percent for White men and 9.6 percent for Asian men [40]. Moreover, job losses during the first half of 2020 occurred disproportionately among Black, Latino, and Asian workers and were especially likely to be experienced by those whose jobs could not be performed at home, most notably frontline workers with little job security and low pay [41, 42]. Another problem with obtaining an accurate description of the labor force is that unemployment rates have fluctuated throughout the period of the pandemic. As a consequence of these issues, our analysis does not account for differential unemployment and labor force participation rates.

As noted above, the analysis includes only individuals who worked in the five years prior to the ACS, the vast majority of whom worked in the year prior to the ACS; we refer to these individuals as either “recent workers” or simply as “workers.” After determining the percent of workers in each racial/ethnic group that hold frontline occupations, we measure the overrepresentation vs. underrepresentation of frontline workers in high-risk occupations. For each of the five risk factors, these estimates denote the proportion of frontline workers in the highest quartile of exposure that are in a given racial/ethnic group relative to the proportion of all frontline workers that are in that racial/ethnic group. We subsequently stratify workers in the highest risk group according to their quartile of our OS measure. The first OS quartile comprises occupations with the lowest standing. Similar to the unstratified estimates, we then consider the proportion of workers in a specified OS quartile (for each of the five risk factors) that are in a given racial/ethnic group relative to the proportion of all frontline workers that are in that group

The descriptive results are presented in five figures: the first includes all 12 racial/ethnic groups, whereas each of the remaining graphs, which are stratified by OS, considers three racial/ethnic groups. All of the figures are heatmaps, with shades of orange indicating overrepresentation of workers in a given risk category relative to all frontline workers (for the particular racial/ethnic group), and shades of green indicating the corresponding underrepresentation in a given risk category. Progressively darker shades reflect successively more extreme values of overrepresentation (orange) or underrepresentation (green). The figures are based on separate calculations for male and female frontline workers.

All calculations use weights provided in the ACS. STATA/MP Version 15.1 was used for the analysis and production of figures [43]. The data underlying these graphs are presented in S1 Appendix.

Results

Table 1 shows the percentages of workers in frontline occupations by sex and race/ethnicity. Employment in frontline occupations varies considerably across groups. For both men and women, Latinos are the most likely workers to hold frontline occupations whereas Chinese and Korean workers are the least likely to do so. Latino, Black, Native American, and Pacific Islander men are the most likely to have frontline occupations–more than 70% of male workers in each of these groups are classified as frontline. In contrast, Vietnamese, Latino, and Filipino women are the most likely female workers to hold frontline occupations. Asian women are at least as likely as Asian men to work in frontline occupations.

Table 1. Number of recent workers and percentage in frontline occupations by race/ethnicity and sex.

	Male		Female
Race/Ethnicity	Number of Recent Workers	Frontline (%)	Number of Recent Workers	Frontline (%)
White	675,542	59.5	632,421	46.2
Black	76,649	72.9	87,168	57.5
Latino	133,868	77.9	120,853	61.8
Chinese	13,095	40.9	14,265	41.3
Filipino	7,798	61.7	10,403	61.2
Vietnamese	5,150	63.5	5,519	66.2
Korean	4,303	44.0	4,684	47.8
Other Asian	22,072	42.1	18,654	57.9
Native American	7,018	75.5	7,059	56.7
Native Hawaiian/Pacific Islander	1,402	73.0	1,353	54.1
Mixed Race	17,555	63.2	17,857	52.0
Other Race	1,695	64.3	1,778	57.3
Total	966,147	63.7	922,014	50.5

Open in a new tab

Note: Data are from the 2018 American Community Survey (ACS). Percentages are calculated using weights provided by the ACS; numbers are unweighted counts of ACS respondents. All racial/ethnic categories except Latino exclude those who reported themselves as Latino/Hispanic in the ACS question on ethnicity.

Fig 1 depicts overrepresentation and underrepresentation of the 12 racial/ethnic groups for each of the five occupational exposures to potential viral transmission, by sex. For example, consider the value of 0.80 for Latino males regarding potential risk of infection (top row, third column of Fig 1). This value suggests that Latinos are underrepresented by 20% in frontline jobs that entail a high risk of infection relative to the proportion of frontline workers that are Latino. It can be calculated from S1 Appendix in the following way: Latino males comprise 17.4% of frontline jobs in the highest quartile of infection risk (5^th column of S1.4) and 21.6% of all frontline workers (5^th column of S1.2); 17.4/21.6 = 0.806 or, allowing for slight round-off error, the value of 0.80 in Fig 1.

Despite elevated COVID-19 mortality among Latino, Black, and Native American populations, Fig 1 shows that these workers are generally underrepresented in high-risk occupations; the main exception is a modest overrepresentation of Black workers in occupations involving risk of infection. In contrast, Whites are modestly overrepresented according to several components of risk, particularly exposure to external customers for men and indoor work for women. The most extreme results are those for Asians, with all ethnic groups of Asian men having more than a twofold overrepresentation in occupations involving indoor work; Filipino men are additionally highly overrepresented in occupations involving high risk of infection.

Fig 1 highlights the complexity of identifying the racial/ethnic groups with the highest potential occupational exposure to the virus. Contrary to expectation based on earlier studies and the media, the values in Fig 1 would suggest that Black and Latino workers do not face higher occupational risks related to viral exposure compared with Whites. However, this conclusion is likely to be incorrect. The classification of high-risk occupations that we have considered thus far does not capture likely variation among occupations in COVID-19 risk reduction measures that are deployed in the workplace and are more likely to be implemented in higher OS occupations. For example, home health aides (relatively low OS and disproportionately held by workers of color) and physicians (high OS and disproportionately White) are both frontline occupations at high risk for infection and physical proximity to other people, but physicians are far more likely to have access to and use PPE, work in frequently sanitized environments, and have had specialized training to reduce potential exposure to infections at work [44, 45].

To get a clearer picture of COVID-19 transmission-related risks faced by workers in different occupations and different racial/ethnic groups, the remaining graphs use OS quartile as a proxy measure for workplace risk mitigation during the COVID-19 pandemic. As discussed above, our use of OS as a proxy measure is based on the assumption that workers in higher OS occupations are more likely to have access to PPE and other mitigation measures compared to workers in lower OS occupations. The most common frontline occupations in the highest OS (4^th) quartile are predominantly in the healthcare sector, most notably registered nurses, physicians, and surgeons. In contrast, the most common frontline occupations in the lowest OS (1^st) quartile include cashiers, drivers, janitors, laborers, stock clerks, and housekeeping cleaners. S2 Appendix lists the most frequently held frontline occupations for each of the five risk indicators, stratified by OS quartile. Only occupations that include at least five percent of workers in the quartile are shown. Although many of the high-risk occupations across all quartiles are in the healthcare sector, numerous occupations outside of the healthcare sector are high-risk on at least one of the five indicators. In the lowest OS quartile, these high-risk occupations include janitors, maids, and other cleaners; carpenters; maintenance and repair workers; food preparation workers and servers; machine operators; stock clerks; and cashiers.

Fig 2, which stratifies the estimates in Fig 1 by occupational standing for White, Black, and Latino workers, provides greater insight into racial/ethnic variation in occupational risks than Fig 1. Among workers in high-risk occupations on each of the five indicators, White workers, both men and women, are generally overrepresented in the 3^rd and especially the 4^th (i.e., highest) OS quartile. Black and Latino workers, on the other hand, are generally overrepresented in the 1^st quartile while being vastly underrepresented (especially Latinos) in the 4^th OS quartile. Latino male workers have the largest overrepresentation (1.5 or 50% greater representation relative to their proportion of all frontline workers) in the lowest OS quartile associated with physical proximity, working, for example, as carpenters, roofers, and food preparation and serving-related workers. Latino female workers have the largest overrepresentation (2.1) in the lowest OS occupations associated with high infection risk, working, for example, as maids and other cleaners.

Figs 3–5 present heatmaps for the remaining racial/ethnic groups, in the same format as Fig 2. As shown in Fig 3, Chinese, Filipino and Vietnamese men, as well as Filipino women, occupy a much larger share of high-risk, but also high OS, occupations relative to their share of frontline occupations, often three to four times as large; this occurs across all five risk indicators. In general, the overrepresentation of men in high-risk jobs with high OS exceeds that for women in these ethnic groups. As shown in the top two panels of Fig 4, a similar pattern occurs for Korean and other Asian workers, i.e., large overrepresentation, especially for men, in high-risk, high OS occupations. The majority of Filipino men and women, and Chinese and Korean women, who are in high OS frontline occupations are registered nurses. The other frequently held occupations for these groups are also in the healthcare sector. Chinese and Korean men with the highest OS frontline occupations are most likely to be physicians and surgeons, but are also employed as dentists, pharmacists and registered nurses. The high proportion of Asians in the healthcare sector largely accounts for their disproportionate presence in high-risk occupations, particularly those involving exposure to infection, close proximity to others, and indoor locations.

Fig 5 — Data are from the 2018 American Community Survey (ACS). Occupations are considered to be high-risk if the average value for the risk factor in the O*NET data falls in the highest quartile of the occupations in the analysis. Occupational standing (OS) is defined as the percentage of ACS respondents reporting this occupation who completed at least one year of college education. The 1^st OS quartile is the lowest quartile and the 4^th (OS) quartile the highest.

These findings for Asian workers contrast strikingly with those for Native Americans in the bottom panel of Fig 4. Native American workers, both men and women, occupy high-risk jobs of the lowest OS quartile more frequently than expected relative to their numbers as frontline workers, particularly for jobs that entail a high risk of infection. Patterns for Pacific Islanders, shown in the top panel of Fig 5, are similar to those of Native Americans. No clear pattern emerges for the remaining heterogeneous racial/ethnic groups considered, mixed race and other race.

Discussion

In this study, we extended earlier work on racial/ethnic differences in potential work-related exposure to COVID-19 by examining 12 racial/ethnic groups and by considering five indicators of potential risk exposure. The analysis is descriptive–portraying potential workplace risks by race, ethnicity and gender–rather than an investigation of the root causes of these patterns. Our central innovation is to use occupational standing (OS) as a proxy for whether workers and their workplaces employ significant COVID-19-related risk reduction strategies (PPE, distancing, sanitation, improved ventilation, etc.)–something not measured in the data we use nor in any other large, nationally representative data sets, to our knowledge. We argue that higher OS workers, even in high-risk occupations, are more likely to have access to (and to use) workplace COVID-19 risk mitigation measures than low OS workers because they are more likely: (a) to work for employers who voluntarily practice risk mitigation and provide PPE and other tools for workers to do so, (b) if necessary, to demand risk reduction measures and to have the bargaining power to obtain them, and (c) to understand (or learn about) COVID-19 transmission routes and to comply with risk reduction strategies or implement their own [26–28].

Once we disaggregate our results by occupational standing, we see large differences in the racial and ethnic distribution of frontline occupations. In contrast to White frontline workers, Latino, Black, and Native American frontline workers are overrepresented in lower OS occupations overall, as well as in lower OS occupations associated with high risk, and thus are probably less likely than Whites to have adequate COVID-19 protections.

In contrast, Asians show more complex patterns, reflecting factors such as the heterogeneity of Asian groups, their immigration history, labor market segregation, and ethnic economic niches. For example, female Filipino workers are disproportionately employed in higher-risk occupations across all of the five risk categories. However, they are much more likely to be in higher OS occupations than other Asian groups, in part because of a history of immigration of Filipino nurses and other health professionals to the US [46].

Although most earlier studies found lower occupational risks for White workers compared with other groups, specific findings vary due largely to the types of occupations examined (e.g., frontline vs. essential) and the occupational characteristics chosen to denote potential risk for COVID-19. Based on data from the 2014–2017 Medical Expenditure Panel Survey and the 2017–2018 American Time Use Survey, Selden and Berdahl found that Black, Latino, and Asian workers were slightly more likely than Whites to work in essential jobs, i.e., at businesses that were allowed to remain open during COVID-19-related shutdowns, although essential White workers were more likely to be able to work from home [23]. Hawkins used data from the 2019 Current Population Survey linked with two measures of risk from O*NET (infection and close proximity) and found an elevated risk for Black workers but virtually none for Latinos for jobs in essential industries [9]. In a report by the Urban Institute based on the 2018 ACS that used O*NET data to identify occupations involving close proximity to other workers, researchers concluded that Black, Latino, and Native American workers were more likely than White workers to have jobs that put them at high risk for viral transmission [16]. Another study, based on employment data for 2014–2019 and only two O*NET measures (working indoors and in close proximity to others), found disproportionate risk exposure for both Black and Latino workers in some occupations [15].

In contrast to much of the earlier work, we focus on frontline occupations (i.e., those in which one-third or fewer workers can feasibly work from home) rather than essential industries. This choice is important because the definition of frontline occupations has remained relatively stable throughout the pandemic and across industries and jurisdictions, whereas occupations classified as essential have varied greatly among states and local jurisdictions and over time. We also expand measurement of potential risk of viral transmission used in previous work that has relied on only one or two occupation-related characteristics. However, our main contribution is that we distinguish among occupations by levels of occupational standing as a proxy for access to and use of COVID-19 risk reduction measures, a strategy that uncovered important differences among racial and ethnic groups that were not apparent in the previous studies.

The racial/ethnic differences by occupational standing highlighted in this paper are consistent with a large literature unrelated to the COVID-19 pandemic. Social science and public health research shows that the history and contemporary effects of racism in the US have led to a labor force highly segregated by race and ethnicity, with Latino and Black workers holding many of the lowest OS and least secure jobs [19, 20, 47, 48]. There is also ample evidence that workers in these occupations were already at higher risk of accidents, injury, infection, and other health problems prior to the pandemic and that these workers also have less control and decision-making ability over how their workplaces are run [49]. These workers may also be more likely to face employer resistance to implementing stricter safety measures that may reduce productivity or profits [47, 50–52]. Thus, a realistic depiction of workplace COVID-19 risks for Latino and Black workers requires consideration of occupational standing and/or careful direct measurement of implementation and regular use of COVID-19 mitigation strategies.

Despite the insights provided by this study, our analysis has several important limitations. First, the 2018 ACS data on recent jobs used here may not provide an accurate picture of employment or occupational distributions during the pandemic. Unemployment rates rose markedly in many areas of the country during the early part of 2020. For example, the rate was 14.7% in April 2020 compared to 3.5% in February 2020 for the US population age 16+ [53]. Unemployment rates were even higher for some racial/ethnic groups, including the Latino (18.9% in April) and Black populations (16.8% in May) [54, 55]. Business closures, layoffs, and unemployment rates have varied a great deal from state to state and over time. Thus, with currently available data it is impossible to determine whether or not workers were employed during 2020 at a particular time and place. Other consequences of the pandemic, such as virtual schooling, have led many parents, especially mothers, to leave the labor force entirely, at least until the pandemic is over [56]. Despite these caveats, we believe that the ACS data provide the best picture currently available of the occupational distribution of the US population at a time close to the onset of the pandemic. A related limitation is that, because our analyses provide no evidence about exposure for people who are unemployed or outside the labor force, our conclusions are necessarily restricted to the employed population.

As is true for all surveys, there is likely to be selective nonresponse in the ACS and misreporting of information (most relevant in our case are misreports of occupational classification and race/ethnicity). Moreover, the ACS occupational classification scheme consisted of a relatively modest number of occupational groups (409 total and 249 frontline), which required us to combine detailed O*NET occupations with potentially varying risk exposures into coarser occupational categories, thereby increasing the amount of intra-occupation variability.

There are also several limitations of the O*NET database and its specific measures we rely on to estimate workplace exposures to COVID-19. Responses to O*NET surveys may be affected by non-response and reporting errors, particularly given the subjectivity of many of the questions, but O*NET does not provide information on differential response rates. As an employer-based survey, O*NET measures the occupational characteristics of workers who are formally employed. Informal workers in our ACS sample, which likely include a substantial portion of immigrants and workers of color, probably have different working conditions than the formally employed workers who provided O*NET data. The omission of informal workers from O*NET may have led us to underestimate the extent of racial and ethnic differences in workplace exposures to COVID-19. In addition, O*NET measures were not designed with a pandemic in mind and thus do not include key information on viral transmission risk. A critical omission for analyses of COVID-19 is data on the implementation of, and worker compliance with, COVID-19 transmission workplace mitigation measures. This lack of information underscores the need to disaggregate workers by occupational standing or another proxy measure for mitigation, as we have done here. There are other limitations to the O*NET data when used in studies of infection-related risks. For example, in the case of contact with others, it would be useful to know the average duration of close contacts, the nature of the contacts and the characteristics of the people with whom the worker interacts. For all of our measures of risk, our analysis would benefit if O*NET provided information on how characteristics of each individual occupation vary by demographic factors (e.g., race/ethnicity, sex, and age of the workers).

Despite these limitations, our results strongly suggest that higher potential work exposures to COVID-19 likely contribute to a higher prevalence of the virus among Latino, Black, and Native American, compared to White, workers in the US. Even though vaccination rates have been increasing, risk reduction, particularly in settings, including workplaces, where transmission is high, must remain an important focus, especially given racial, ethnic, and social class disparities in vaccination rates and the emergence of newer, more transmissible SAR-CoV2 variants.

Reducing potential exposure for workers in low OS occupations requires multiple coordinated measures beyond PPE. Carlsten and colleagues draw on industrial hygiene’s hierarchy of controls framework to outline three types of measures employers should take to mitigate COVID-19 risks for their workers [17]. The first, eliminating exposure to the SARS CoV-2 virus completely, would clearly be the most effective measure–via options such as requiring vaccination, symptom and illness reporting, COVID-19 testing, providing paid sick leave that is ample and accessible, and providing telecommuting and work-from-home options. Second, engineering and administrative controls (e.g., physical barriers between people, improved ventilation systems, regular disinfecting protocols, and staggered work schedules) are the next most effective. Finally, provision and use of personal protective equipment (e.g., masks, shields) is the least effective, albeit essential (particularly combined with engineering and administrative controls) if eliminating exposure is not possible [17]. Analysis of survey data from June 2020 for non-health care workers who were unable to work remotely showed that voluntary use of COVID-19 workplace hazard controls (defined as PPE and other physical barriers) was approximately double when employers provided these hazard controls than when they did not [29]. The differences were particularly large among low-income workers, perhaps because they were less likely to be able to afford PPE themselves. All of these measures would additionally benefit the broader community through reduced transmission.

From a public policy point of view, the question is how to ensure that employers protect workers by implementing these risk mitigation strategies, particularly in workplaces employing lower OS workers. In the US, employers are legally required to provide a workplace free of recognized health and safety hazards. The federal Occupational Safety and Health Administration (OSHA), the Centers for Disease Control and Prevention (CDC) and state governmental agencies have provided guidance to employers about exposure controls against COVID-19 transmission [57], but enforcement has been limited [58]. Given the importance of workplaces as venues for transmission of COVID-19 and pathogens in future pandemics, federal and state agencies regulating worker safety need better tools and greater emergency power to act on behalf of worker health.

Supporting information

S1 Appendix. Racial/ethnic distributions within each occupational subset.

(DOCX)

Click here for additional data file.^{(68.4KB, docx)}

S2 Appendix. Common high-risk occupations by occupational standing quartile.

(DOCX)

Click here for additional data file.^{(34.6KB, docx)}

Acknowledgments

The authors would like to thank Francine Blau at Cornell University, Josefine Koebe at Universität Hamburg, and Pamela Meyerhofer at Montana State University for sharing their code to identify frontline occupations. We also thank Dana Glei and the anonymous reviewers for suggestions on earlier drafts of this paper.

Data Availability

All data used in this analysis are publicly available. IPUMS USA Version 10.0 is available at https://doi.org/10.18128/D010.V10.0 and O*NET data are available at https://www.onetcenter.org/db_releases.html.

Funding Statement

Research reported in this publication was supported by the National Institute on Aging through grant numbers R01AG061094 (UCLA & Princeton, ARP and NG) and T32AG000037 (TA) and the Eunice Kennedy Shriver National Institute on Child Health and Human Development through grant number P2CHD041022 (ARP). The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.

References

1.Andrasfay T, Goldman N. Reductions in 2020 US life expectancy due to COVID-19 and the disproportionate impact on the Black and Latino populations. Proc Natl Acad Sci. 2021;118. doi: 10.1073/pnas.2014746118 [DOI] [PMC free article] [PubMed] [Google Scholar]
2.Andrasfay T, Goldman N. Association of the COVID-19 Pandemic With Estimated Life Expectancy by Race/Ethnicity in the United States, 2020. JAMA Netw Open. 2021;4: e2114520–e2114520. doi: 10.1001/jamanetworkopen.2021.14520 [DOI] [PMC free article] [PubMed] [Google Scholar]
3.Leggat-Barr K, Uchikoshi F, Goldman N. COVID-19 Risk Factors and Mortality among Native Americans. medRxiv. 2021; 2021.03.13.21253515. doi: 10.1101/2021.03.13.21253515 [DOI] [Google Scholar]
4.Rentsch CT, Kidwai-Khan F, Tate JP, Park LS, Jr JTK, Skanderson M, et al. Patterns of COVID-19 testing and mortality by race and ethnicity among United States veterans: A nationwide cohort study. PLOS Med. 2020;17: e1003379. doi: 10.1371/journal.pmed.1003379 [DOI] [PMC free article] [PubMed] [Google Scholar]
5.Sáenz R, Garcia MA. The Disproportionate Impact of COVID-19 on Older Latino Mortality: The Rapidly Diminishing Latino Paradox. J Gerontol Ser B. 2020;XX: 1–7. doi: 10.1093/geronb/gbaa158 [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Wang D, Gee GC, Bahiru E, Yang EH, Hsu JJ. Asian-Americans and Pacific Islanders in COVID-19: Emerging Disparities Amid Discrimination. J Gen Intern Med. 2020. [cited 6 Nov 2020]. doi: 10.1007/s11606-020-06264-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Woolf SH, Masters RK, Aron LY. Effect of the covid-19 pandemic in 2020 on life expectancy across populations in the USA and other high income countries: simulations of provisional mortality data. BMJ. 2021;373: n1343. doi: 10.1136/bmj.n1343 [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Casura L, Lowe R Jr., Martinez C, Serpas S, Castellanos V, Singelmann J. Frontline workers in the U.S.: race, ethnicity, and gender. International Union for the Scientific Study of Population; 2020Sep. Available: https://www.niussp.org/article/frontline-workers-in-the-u-s-race-ethnicity/ [Google Scholar]
9.Hawkins D. Differential occupational risk for COVID-19 and other infection exposure according to race and ethnicity. Am J Ind Med. 2020;65. doi: 10.1002/ajim.23145 [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Covid-19 Deaths Skew Younger Among Minorities—WSJ. The Wall Street Journal. 17 Aug 2020. Available: https://www-wsj-com.libproxy1.usc.edu/articles/covid-19-deaths-strike-early-for-many-minorities-11597582801. Accessed 10 Nov 2020.
11.Oppel Jr RA, Gebeloff R, Lai KKR, Wright W, Smith M. The Fullest Look Yet at the Racial Inequity of Coronavirus. The New York Times. 5 Jul 2020. Available: https://www.nytimes.com/interactive/2020/07/05/us/coronavirus-latinos-african-americans-cdc-data.html. Accessed 10 Nov 2020.
12.Reckdahl K. “The impact of work”: On-the-job coronavirus exposure a key driver in Black, Latino communities. NOLA.com. 17 Oct 2020. Available: https://www.nola.com/news/business/article_d0e20ff2-0a62-11eb-8314-af3dffbffd15.html. Accessed 10 Nov 2020.
13.U.S. News and World Report. AHA News: Hispanics Are Paying the Price for Being “Essential” During the Pandemic. US News & World Report. 9 Oct 2020. Available: https://www.usnews.com/news/health-news/articles/2020-10-09/aha-news-hispanics-are-paying-the-price-for-being-essential-during-the-pandemic. Accessed 10 Nov 2020.
14.Erdman SL. More Hispanic workers impacted by Covid-19 in food processing and agriculture workplaces, CDC study finds. CNN. 20 Oct 2020. Available: https://www.cnn.com/2020/10/20/health/cdc-study-covid-hispanic-workers/index.html. Accessed 10 Nov 2020.
15.Cox-Ganser JM, Henneberger PK. Occupations by Proximity and Indoor/Outdoor Work: Relevance to COVID-19 in All Workers and Black/Hispanic Workers. Am J Prev Med. 2021. [cited 30 Mar 2021]. doi: 10.1016/j.amepre.2020.12.016 [DOI] [PMC free article] [PubMed] [Google Scholar]
16.Dubay L, Aarons J, Brown KS, Kenney GM. How Risk of Exposure to the Coronavirus at Work Varies by Race and Ethnicity and How to Protect the Health and Well-Being of Workers and Their Families. Urban Institute; 2020Dec. Available: https://www.urban.org/research/publication/how-risk-exposure-coronavirus-work-varies-race-and-ethnicity-and-how-protect-health-and-well-being-workers-and-their-families/view/full_report [Google Scholar]
17.Carlsten C, Gulati M, Hines S, Rose C, Scott K, Tarlo SM, et al. COVID-19 as an occupational disease. Am J Ind Med. 2021;64: 227–237. 10.1002/ajim.23222 [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Hawkins D, Davis L, Kriebel D. COVID-19 deaths by occupation, Massachusetts, March 1–July 31, 2020. Am J Ind Med. 2021;64: 238–244. doi: 10.1002/ajim.23227 [DOI] [PMC free article] [PubMed] [Google Scholar]
19.Bailey ZD, Krieger N, Agénor M, Graves J, Linos N, Bassett MT. Structural racism and health inequities in the USA: evidence and interventions. The Lancet. 2017;389: 1453–1463. doi: 10.1016/S0140-6736(17)30569-X [DOI] [PubMed] [Google Scholar]
20.del Río C, Alonso-Villar O. The Evolution of Occupational Segregation in the United States, 1940–2010: Gains and Losses of Gender–Race/Ethnicity Groups. Demography. 2015;52: 967–988. doi: 10.1007/s13524-015-0390-5 [DOI] [PubMed] [Google Scholar]
21.Chetty R, Hendren N, Jones MR, Porter SR. Race and Economic Opportunity in the United States: an Intergenerational Perspective*. Q J Econ. 2020;135: 711–783. doi: 10.1093/qje/qjz042 [DOI] [Google Scholar]
22.Pager D, Shepherd H. The Sociology of Discrimination: Racial Discrimination in Employment, Housing, Credit, and Consumer Markets. Annu Rev Sociol. 2008;34: 181–209. doi: 10.1146/annurev.soc.33.040406.131740 [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Selden TM, Berdahl TA. COVID-19 And Racial/Ethnic Disparities In Health Risk, Employment, And Household Composition. Health Aff (Millwood). 2020;39: 1624–1632. doi: 10.1377/hlthaff.2020.00897 [DOI] [PubMed] [Google Scholar]
24.Blau FD, Koebe J, Meyerhofer PA. Who are the Essential and Frontline Workers? National Bureau of Economic Research; 2020 Sep. Report No.: w27791. doi: 10.3386/w27791 [DOI] [PMC free article] [PubMed]
25.Chen Y-H, Glymour M, Riley A, Balmes J, Duchowny K, Harrison R, et al. Excess mortality associated with the COVID-19 pandemic among Californians 18–65 years of age, by occupational sector and occupation: March through October 2020. medRxiv. 2021; 2021.01.21.21250266. doi: 10.1101/2021.01.21.21250266 [DOI] [PMC free article] [PubMed] [Google Scholar]
26.Phelan JC, Link BG, Tehranifar P. Social Conditions as Fundamental Causes of Health Inequalities: Theory, Evidence, and Policy Implications. J Health Soc Behav. 2010;51: S28–S40. doi: 10.1177/0022146510383498 [DOI] [PubMed] [Google Scholar]
27.Gaitens J, Condon M, Fernandes E, McDiarmid M. COVID-19 and Essential Workers: A Narrative Review of Health Outcomes and Moral Injury. Int J Environ Res Public Health. 2021;18: 1446. doi: 10.3390/ijerph18041446 [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Billock RM, Groenewold MR, Free H, Haring Sweeney M, Luckhaupt SE. Required and Voluntary Occupational Use of Hazard Controls for COVID-19 Prevention in Non–Health Care Workplaces—United States, June 2020. MMWR Morb Mortal Wkly Rep. 2021;70. doi: 10.15585/mmwr.mm7007a5 [DOI] [PMC free article] [PubMed] [Google Scholar]
29.Hoven H, Wahrendorf M, Siegrist J. Occupational position, work stress and depressive symptoms: a pathway analysis of longitudinal SHARE data. J Epidemiol Community Health. 2015;69: 447–452. doi: 10.1136/jech-2014-205206 [DOI] [PMC free article] [PubMed] [Google Scholar]
30.Landsbergis PA, Grzywacz JG, LaMontagne AD. Work organization, job insecurity, and occupational health disparities. Am J Ind Med. 2014;57: 495–515. doi: 10.1002/ajim.22126 [DOI] [PubMed] [Google Scholar]
31.Clougherty JE, Souza K, Cullen MR. Work and its role in shaping the social gradient in health. Ann N Y Acad Sci. 2010;1186: 102–124. doi: 10.1111/j.1749-6632.2009.05338.x [DOI] [PMC free article] [PubMed] [Google Scholar]
32.Ndugga N, Pham O, Hill L, Artiga S, Parker N. Latest Data on COVID-19 Vaccinations by Race/Ethnicity. San Francisco, CA: Kaiser Family Foundation; 2021Jun. Available: https://www.kff.org/coronavirus-covid-19/issue-brief/latest-data-on-covid-19-vaccinations-race-ethnicity/ [Google Scholar]
33.Ruggles S, Flood S, Goeken R, Grover J, Meyer E, Pacas J, et al. IPUMS USA: Version 10.0 [dataset]. Minneapolis, MN: IPUMS; 2020. Available: 10.18128/D010.V10.0 [DOI] [Google Scholar]
34.O*NET OnLine. National Center for O*NET Development; Available: https://www.onetonline.org/
35.Wilson N, Corbett S, Tovey E. Airborne transmission of covid-19. BMJ. 2020;370. doi: 10.1136/bmj.m3206 [DOI] [PubMed] [Google Scholar]
36.Hauser RM, Warren JR. Socioeconomic Indexes for Occupations: A Review, Update, and Critique: Sociol Methodol. 2016. [cited 6 Nov 2020]. Available: https://journals.sagepub.com/doi/10.1111/1467-9531.271028 [Google Scholar]
37.Tomer A, Kane JW. How to protect essential workers during COVID-19. Washington, D.C.: The Brookings Institution; 2020Mar. Available: https://www.brookings.edu/research/how-to-protect-essential-workers-during-covid-19/ [Google Scholar]
38.Rogers TN, Rogers CR, VanSant-Webb E, Gu LY, Yan B, Qeadan F. Racial Disparities in COVID-19 Mortality among Essential Workers in the United States. World Med Health Policy. 2020; 10.1002/wmh3.358. doi: 10.1002/wmh3.358 [DOI] [PMC free article] [PubMed] [Google Scholar]
39.Dingel JI, Neiman B. How Many Jobs Can be Done at Home? National Bureau of Economic Research; 2020 Apr. Report No.: w26948. doi: 10.3386/w26948 [DOI]
40.Unemployment rates by age, sex, race, and Hispanic or Latino ethnicity. Washington, DC: U.S. Bureau of Labor Statistics; 2020 Oct. Report No.: E-16.
41.Gemelas J, Davison J, Keltner C, Ing S. Inequities in Employment by Race, Ethnicity, and Sector During COVID-19. J Racial Ethn Health Disparities. 2021. [cited 30 Mar 2021]. doi: 10.1007/s40615-021-00963-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
42.Baker MG. Nonrelocatable Occupations at Increased Risk During Pandemics: United States, 2018. Am J Public Health. 2020;110: 1126–1132. doi: 10.2105/AJPH.2020.305738 [DOI] [PMC free article] [PubMed] [Google Scholar]
43.StataCorp. Stata Statistical Software: Release 15. College Station, TX:StataCorp LLC.: StataCrop; 2017. [Google Scholar]
44.Sterling MR, Tseng E, Poon A, Cho J, Avgar AC, Kern LM, et al. Experiences of Home Health Care Workers in New York City During the Coronavirus Disease 2019 Pandemic: A Qualitative Analysis. JAMA Intern Med. 2020;180: 1453–1459. doi: 10.1001/jamainternmed.2020.3930 [DOI] [PMC free article] [PubMed] [Google Scholar]
45.Shang J, Chastain AM, Perera UGE, Dick AW, Fu CJ, Madigan EA, et al. The state of infection prevention and control at home health agencies in the United States prior to COVID-19: A cross-sectional study. Int J Nurs Stud. 2021;115: 103841. doi: 10.1016/j.ijnurstu.2020.103841 [DOI] [PMC free article] [PubMed] [Google Scholar]
46.Choy CC. Empire of Care: Nursing and Migration in Filipino American History. Duke University Press; 2003. [Google Scholar]
47.Hall M, Greenman E. The Occupational Cost of Being Illegal in the United States: Legal Status, Job Hazards, and Compensating Differentials. Int Migr Rev. 2015;49: 406–442. doi: 10.1111/imre.12090 [DOI] [PMC free article] [PubMed] [Google Scholar]
48.Wilson FD. Generational changes in racial inequality in occupational attainment, 1950–2010: A synthetic cohort analysis. Bois Rev. 2017;14: 387–425. doi: 10.1017/S1742058X17000170 [DOI] [Google Scholar]
49.Steege AL, Baron SL, Marsh SM, Menéndez CC, Myers JR. Examining occupational health and safety disparities using national data: A cause for continuing concern. Am J Ind Med. 2014;57: 527–538. doi: 10.1002/ajim.22297 [DOI] [PMC free article] [PubMed] [Google Scholar]
50.Krieger N, Chen JT, Waterman PD, Hartman C, Stoddard AM, Quinn MM, et al. The inverse hazard law: Blood pressure, sexual harassment, racial discrimination, workplace abuse and occupational exposures in US low-income black, white and Latino workers. Soc Sci Med. 2008;67: 1970–1981. doi: 10.1016/j.socscimed.2008.09.039 [DOI] [PubMed] [Google Scholar]
51.Quandt SA, LaMonto NJ, Mora DC, Talton JW, Laurienti PJ, Arcury TA. COVID-19 Pandemic among Latinx Farmworker and Nonfarmworker Families in North Carolina: Knowledge, Risk Perceptions, and Preventive Behaviors. Int J Environ Res Public Health. 2020;17: 5786. doi: 10.3390/ijerph17165786 [DOI] [PMC free article] [PubMed] [Google Scholar]
52.McClure ES, Vasudevan P, Bailey Z, Patel S, Robinson WR. Racial Capitalism Within Public Health—How Occupational Settings Drive COVID-19 Disparities. Am J Epidemiol. 2020;189: 1244–1253. doi: 10.1093/aje/kwaa126 [DOI] [PMC free article] [PubMed] [Google Scholar]
53.U.S. Bureau of Labor Statistics. Unemployment Rate [UNRATE], retrieved from FRED. Federal Reserve Bank of St. Louis; 2020 Nov. Available: https://fred.stlouisfed.org/series/UNRATE
54.U.S. Bureau of Labor Statistics. Unemployment Rate—Hispanic of Latino [LNS14000009], retrieved from FRED. Federal Reserve Bank of St. Louis; 2020 Nov. Available: https://fred.stlouisfed.org/series/LNS14000009
55.U.S. Bureau of Labor Statistics. Unemployment Rate—Black or African American [LNS14000006], retrieved from FRED. Federal Reserve Bank of St. Louis; 2020 Nov. Available: https://fred.stlouisfed.org/series/LNS14000006
56.Kochhar R. Fewer mothers and fathers in U.S. are working due to COVID-19 downturn; those at work have cut hours. Washington, D.C.: Pew Research Center; 2020Oct. Available: https://www.pewresearch.org/fact-tank/2020/10/22/fewer-mothers-and-fathers-in-u-s-are-working-due-to-covid-19-downturn-those-at-work-have-cut-hours/ [Google Scholar]
57.Protecting Workers: Guidance on Mitigating and Preventing the Spread of COVID-19 in the Workplace | Occupational Safety and Health Administration. Washington, DC: Occupational Safety & Health Administration; 2021Jan. Available: https://www.osha.gov/coronavirus/safework [Google Scholar]
58.Michaels D, Wagner GR. Occupational Safety and Health Administration (OSHA) and Worker Safety During the COVID-19 Pandemic. JAMA. 2020;324: 1389–1390. doi: 10.1001/jama.2020.16343 [DOI] [PubMed] [Google Scholar]

PLoS One. doi: 10.1371/journal.pone.0256085.r001

Decision Letter 0

Marlene Camacho-Rivera

28 Jan 2021

PONE-D-20-36345

Racial and Ethnic DIfferentials in COVID-19-Related Job Exposures by Occupational Status in the US

PLOS ONE

Dear Dr. Goldman,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript within 90 days of receipt of this email. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.
A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.
An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

We look forward to receiving your revised manuscript.

Kind regards,

Marlene Camacho-Rivera, ScD, MPH

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: N/A

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The manuscript is very well written. The use of “occupational status” is innovative and informative. The manuscript has many strengths and few minor weaknesses. The findings provide more evidence on the racial disparities experienced by workers of color during COVID-19 pandemic for policy considerations.

I just have a few suggestions:

- Typically, “occupational status” is defined by a combination of educational attainment and income. In this article, “occupational status” is defined by educational attainment only. Please comment on this point in your limitation section and whether your definition might or might not change your overall interpretation.

- Did you not have information on income?

- The text in the figures are blurry and hard to read.

References:

https://macses.ucsf.edu/research/socialenviron/occupation.php

Reviewer #2: Language tweak throughout: use Black workers instead of Blacks

Introduction: since the submission of this paper, there have been a few key studies about occupational risks by race/ethnicity and by industry. Suggest updating the lit review.

Introduction could be a lot meatier and get into some of the systemic drivers of potential reasons for greater exposure risk among low-wage and racial/ethnic minority workers. There’s been a ton of great work here—again, much of it published since this paper was submitted. But also the analysis needs more context and some data on, for example, disproportionate risks of hospitalizations and deaths by race and ethnicity, and hypothesized reasons. It also needs more citation in general, for statements like that in lines 85-87.

“Occupational status” is a very confusing term, and in reading the abstract I initially thought it was a typo. I assumed that it meant whether one was working or not working (synonymous with "employment status"), as it’s commonly used in that context in the occupational health literature. However, I see that you’re using it to mean “status” as in the social hierarchy, rather than status as in present/not present. I would suggest using a different term throughout.

In the intro, the authors appear to conflate race and ethnicity with “occupational status.” (lines 91-92). While the two often travel together, at this point in the manuscript they should be treated as separate unless they back up their assertions with the citations that appear later in the discussion.

Does the ACS include both occupation and industry for individual workers? When linking with o*net, do you consider industry, or just occupation? Industry could be quite important for establishing baseline level of risk by occupational setting. For example, a janitor in a hospital would be much more highly exposed than a janitor in a warehouse.

For “frontline status,” how do you account for the differential job loss that industries and occupations experienced during the pandemic? Yes, a server in a restaurant would be considered frontline because it can’t be done from home, but almost all servers lost their jobs—reducing their exposure status in the process. Not accounting for potential job loss could lead to substantial exposure misclassification

Table 1: suggest ordering by either size of the population or % frontline (for men) to make it easier to find the relevant information

The figures are nearly impossible to read, much less interpret. I printed out the manuscript and was unable to see the differences between the groups. In addition, readers are not good at comparing the relative size of bars on a figure aside from the categories at the beginning and the end, and it requires a heavy cognitive lift on the part of the reader. Strongly consider re-imagining how this data could be better visualized; I suggest any of Stephanie Evergreen’s books.

The point you make in lines 276-277 is precisely why industry, in addition to occupation, is crucial to consider. While educational attainment is also important, industry is available at an individual level and so there is lower risk of misclassification bias. From the discussion I understand the reasons that you chose to rely on occupation rather than industry (murky definitions of essential workers), but could you not use the combination of industry/occupation for the exposure assessment, and only industry for the comparison?

Re: the limitation of women being forced out of the labor market due to having young children, I imagine that the ACS measures both child age and marital status. How could the authors leverage this information to do a sensitivity analysis?

In the paragraph beginning on line 433, there’s a highly cited framework called the NIOSH Hierarchy of Controls that refers to different strategies for risk reduction/mitigation. Suggest citing it here and using it as an organizing framework for this paragraph/section.

Finally, any analysis that looks at racial disparities should explicitly discuss systemic racism as an upstream driver of the distribution of race/ethnicity by occupations.

Overall, this manuscript is very much in need of a coauthor or collaborator with expertise in occupational epidemiology, especially when using a job exposure matrix such as onet. There are clear gaps in knowledge (use of o*net, hierarchy of controls, intersection of occupation and industry classifications, testing for precision and accuracy of the assessed exposure) that should be addressed in a paper that explicitly addresses occupational exposure burden. I also suggest adding a social epidemiology coauthor or collaborator with experience in applying health disparities frameworks.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

PLoS One. 2021 Sep 1;16(9):e0256085. doi: 10.1371/journal.pone.0256085.r002

Author response to Decision Letter 0

6 Apr 2021

We thank both reviewers for their helpful comments that we believe strengthened our paper. Below we present each of the comments (in italics) followed by our response.

Reviewer #1

• The manuscript is very well written. The use of “occupational status” is innovative and informative. The manuscript has many strengths and few minor weaknesses. The findings provide more evidence on the racial disparities experienced by workers of color during COVID-19 pandemic for policy considerations.

We thank the reviewer for this positive feedback.

• I just have a few suggestions:

Typically, “occupational status” is defined by a combination of educational attainment and income. In this article, “occupational status” is defined by educational attainment only. Please comment on this point in your limitation section and whether your definition might or might not change your overall interpretation. Did you not have information on income?

Duncan’s socioeconomic status index (SEI) does indeed combine occupational education and occupational status into a weighted index. However, Hauser and Warren (1997) – cited in the paper and below – argued persuasively that occupational education and occupational income should be used as two separate measures because they have different effects and much of the literature now does so. Our goal was to identify a simple proxy for the amount of power and control workers are likely to have in determining how they do their work and whether they have access to PPE and other workplace protections. We do have information on income, but we chose occupational education alone for this purpose to avoid complicating our results further by using two SES separate indices. Occupational education is easy to understand, frequently used in the sociological literature, and, compared to occupational income, less likely to have been altered by the pandemic. Furthermore, education is typically completed at young ages, whereas job choices (often based heavily on wage and salary rates) are made throughout the working years – so occupational income is more endogenous to the choice of job than is occupational education.

We have now changed the term we use from “occupational status” to “occupational standing” to avoid confusion.

We have revised our discussion of this issue in the text.

Reference: Hauser RM, Warren JR. (1997) Socioeconomic Indexes for Occupations: A Review, Update, and Critique: Sociological Methodology. Available: https://journals.sagepub.com/doi/10.1111/1467-9531.271028

• The text in the figures are blurry and hard to read.

We have completely revamped the figures. The figures now directly examine the overrepresentation vs. underrepresentation of frontline workers in high-risk occupations and show the results both numerically and by color (using a heatmap). These graphs are much easier to read and to interpret than the previous figures.

Reviewer #2

• Language tweak throughout: use Black workers instead of Blacks

Thanks for catching this. We have fixed this throughout the paper.

• Introduction: since the submission of this paper, there have been a few key studies about occupational risks by race/ethnicity and by industry. Suggest updating the lit review.

We have updated the literature review (primarily in the Introduction and Discussion sections), and now include studies published through March 2021 .

• Introduction could be a lot meatier and get into some of the systemic drivers of potential reasons for greater exposure risk among low-wage and racial/ethnic minority workers. There’s been a ton of great work here—again, much of it published since this paper was submitted. But also the analysis needs more context and some data on, for example, disproportionate risks of hospitalizations and deaths by race and ethnicity, and hypothesized reasons.

It also needs more citation in general, for statements like that in lines 85-87.

As noted above, we have revised the introduction to include new papers. We have also added language that refers to the “the long history and contemporary effects of structural racism on occupational segregation in the US” and provided references.

We agree that we could have written a much broader introduction discussing the role of structural racism and other social processes in producing the racial and ethnic distribution of workers by occupation, the reasons that lower wage workers are less likely to have access to PPE, and the disproportionate risks of hospitalizations and deaths. These are issues that we have written about at length in other work. However, this paper is specifically focused on the empirical evidence about the potential exposure to COVID-19 of frontline workers in different racial and ethnic groups. There has been extensive discussion in the press and academic publications about the issue, but limited empirical evidence – particularly evidence that includes proxy information on the likelihood of risk reduction mitigation, as we try to provide in this paper. We believe that it is essential to make these key empirical findings available to as broad an audience as possible as quickly as possible and, therefore, we have kept the narrow focus.

• “Occupational status” is a very confusing term, and in reading the abstract I initially thought it was a typo. I assumed that it meant whether one was working or not working (synonymous with "employment status"), as it’s commonly used in that context in the occupational health literature. However, I see that you’re using it to mean “status” as in the social hierarchy, rather than status as in present/not present. I would suggest using a different term throughout.

At your suggestion, we have changed the term to “occupational standing” throughout the paper, figures and appendices. “Occupational status” is a term from the sociological and social science literature used to refer to an individual’s position in the occupational hierarchy. However, we agree that it can be a confusing term for readers in other fields and so we have made the change.

• In the intro, the authors appear to conflate race and ethnicity with “occupational status.” (lines 91-92). While the two often travel together, at this point in the manuscript they should be treated as separate unless they back up their assertions with the citations that appear later in the discussion.

The lines referred to in the reviewer’s comment read:

91 Examining results by occupational status leads to a clearer picture of racial/ethnic differentials in

92 potential exposure to COVID-19 transmission.

Note that “occupational status” originally used in line 91 is now “occupational standing.” Our point was that differences in the distribution of occupations by race and ethnicity are likely to contribute to differences in exposure to COVID-19, since Black, Latino, and some other groups of workers are more likely to work in jobs without adequate protection compared to whites and some other groups. We are not in any way conflating race and ethnicity with occupational (or socioeconomic) status. The sentence in lines 91-92 was proceeded by the following text, which we believe makes the meaning of the text in 91-92 clear:

“In contrast, higher status workers have better access to risk mitigation measures such as personal protective equipment (PPE), frequent sanitation, enforced distancing, partitions, better ventilation, and new filtration systems.”

• Does the ACS include both occupation and industry for individual workers? When linking with o*net, do you consider industry, or just occupation? Industry could be quite important for establishing baseline level of risk by occupational setting. For example, a janitor in a hospital would be much more highly exposed than a janitor in a warehouse.

We agree that use of PPE and other forms of protection are likely to vary by industry as well as occupation. The ACS data contains both variables. However, the O*NET data provides exposures only for occupations (see https://www.onetcenter.org/overview.html). Thus, even if we subdivide ACS respondents both by occupation and industry, the exposure information we obtain from O*NET would be identical for respondents in the same occupation, regardless of industry.

• For “frontline status,” how do you account for the differential job loss that industries and occupations experienced during the pandemic? Yes, a server in a restaurant would be considered frontline because it can’t be done from home, but almost all servers lost their jobs—reducing their exposure status in the process. Not accounting for potential job loss could lead to substantial exposure misclassification

We agree that this is a potential problem, in our analysis and in all those which have sought to use pre-pandemic data to examine potential effects of the pandemic. The issue is even more complicated because “job loss” for individual workers changed a great deal during the pandemic depending on what was and was not open. In many states, some businesses were closed, then reopened for a while, then closed again, then reopened, etc. Many of these business that were unable to keep workers on the payroll while closed (although the Payroll Protection Program initially helped a lot of employers continue to pay workers) reemployed them when they reopened. The detailed data to take this issue into account is not yet available. However, we added a sentence to the text that refers to differences by race and ethnicity in job losses during the pandemic, and we specifically acknowledge our inability to account for job losses and unemployment.

With regard to the broader point about exposure misclassification, throughout the text we are now careful not to imply that these measures correspond to an individual’s experienced risk during the pandemic; rather, they correspond to racial and ethnic patterns of employment in high-risk occupations.

• Table 1: suggest ordering by either size of the population or % frontline (for men) to make it easier to find the relevant information

We believe that it is important to keep the order of racial/ethnic groups consistent with the graphs (and appendix tables) as well as to have the groups that have received the most attention listed first (White, Black and Latino) and to keep the four Asian groups together. We have changed the placement of Native Americans to come after the four Asian groups and directly before Asian and Pacific Islanders. This ordering is now consistent in Table 1, the figures, and the appendices.

• The figures are nearly impossible to read, much less interpret. I printed out the manuscript and was unable to see the differences between the groups. In addition, readers are not good at comparing the relative size of bars on a figure aside from the categories at the beginning and the end, and it requires a heavy cognitive lift on the part of the reader. Strongly consider re-imagining how this data could be better visualized; I suggest any of Stephanie Evergreen’s books.

As we noted in our response to Reviewer 1, we have completely redone the figures. The reader no longer is asked to assess the relative size of bars (an admittedly difficult and unreasonable task); these relative measures are computed directly and are shown as measures of overrepresentation and underrepresentation in various categories of occupations. Interpretation is aided by the use of a heatmap and a detailed numerical example.

• The point you make in lines 276-277 is precisely why industry, in addition to occupation, is crucial to consider. While educational attainment is also important, industry is available at an individual level and so there is lower risk of misclassification bias. From the discussion I understand the reasons that you chose to rely on occupation rather than industry (murky definitions of essential workers), but could you not use the combination of industry/occupation for the exposure assessment, and only industry for the comparison?

The lines referenced (276-277) have been rewritten to better clarify our focus on occupational standing and avoid comparing across industry. We now provide examples of two occupations within the healthcare sector—home health aides and physicians—to emphasize how occupations with similar levels of risk factors according to O*NET can have very different statuses. We have also included some citations to recent work to support this claim that lower status healthcare workers have received inadequate training in infectious disease protocol and faced severe shortages of personal protective equipment.

As noted above, only occupation-related exposure information is available from O*NET so we are unable to directly consider how these risk factors vary across industry.

• Re: the limitation of women being forced out of the labor market due to having young children, I imagine that the ACS measures both child age and marital status. How could the authors leverage this information to do a sensitivity analysis?

An analysis of women who had to leave the labor market to take care of children when schools were closed is very important, but is beyond the scope of this paper. From media and emerging research, we know that parents have made a huge variety of arrangements for children during this period, which have been affected by the type of child care (or after school care) arrangements they had pre-pandemic, the availability of family close by, services offered by community organizations, etc. With our available data, we do not have the ability to consider all the relevant factors necessary to make predictions about which women were likely to exit the labor force.

• In the paragraph beginning on line 433, there’s a highly cited framework called the NIOSH Hierarchy of Controls that refers to different strategies for risk reduction/mitigation. Suggest citing it here and using it as an organizing framework for this paragraph/section.

This part of the discussion section is now rewritten and refers both to the HoC and to a thoughtful article applying this framework to COVID-19 hazard mitigation.

• Finally, any analysis that looks at racial disparities should explicitly discuss systemic racism as an upstream driver of the distribution of race/ethnicity by occupations.

Concerns about systemic racism and its consequences for workers’ location in the employment hierarchy and for health led us to undertake this research. Our research team has a long track record of research on racism, discrimination, and social inequality and health outcomes. However, in this paper, in order to complete and publish the research while it still can be useful to readers, we decided to focus on the effects of the occupational distribution of workers by race and ethnicity on exposure rather than on the causes of the occupational distribution of workers by race and ethnicity itself.

While we note the role of racism in occupational segregation, providing an adequate treatment of the social processes, including systemic racism, which have led to current occupational disparities by race and ethnicity would require a considerable lengthening of the paper and additional empirical analyses that would delay the availability of these results and greatly enlarge the scope of the paper.

• Overall, this manuscript is very much in need of a coauthor or collaborator with expertise in occupational epidemiology, especially when using a job exposure matrix such as onet. There are clear gaps in knowledge (use of o*net, hierarchy of controls, intersection of occupation and industry classifications, testing for precision and accuracy of the assessed exposure) that should be addressed in a paper that explicitly addresses occupational exposure burden. I also suggest adding a social epidemiology coauthor or collaborator with experience in applying health disparities frameworks.

The authors of this paper comprise a multidisciplinary research team whose members include epidemiologists and social scientists with an extensive track record of research on health disparities by race, ethnicity, and immigration status and the social causes of occupational attainment and social position. We also have considerable experience with analyses of O*NET, occupational and industry classifications, and occupational exposures. As noted above, this paper is intentionally sharply focused because, given our experience and background, we believe that it is essential to provide solid empirical results on the extreme disadvantage in terms of exposure to COVID-19 faced by many frontline workers during the pandemic, particularly among Black, Latino, Native American, and some other groups of workers.

Attachment

Submitted filename: Response to Reviewers.docx

Click here for additional data file.^{(23.3KB, docx)}

PLoS One. doi: 10.1371/journal.pone.0256085.r003

Decision Letter 1

Marlene Camacho-Rivera

21 Jun 2021

PONE-D-20-36345R1

Racial and Ethnic DIfferentials in COVID-19-Related Job Exposures by Occupational Standing in the US

PLOS ONE

Dear Dr. Goldman,

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.
A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.
An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Marlene Camacho-Rivera, ScD, MPH

Academic Editor

PLOS ONE

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #3: (No Response)

Reviewer #4: (No Response)

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

Reviewer #1: Yes

Reviewer #3: Partly

Reviewer #4: Partly

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #3: I Don't Know

Reviewer #4: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

Reviewer #1: Yes

Reviewer #3: Yes

Reviewer #4: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

Reviewer #1: Yes

Reviewer #3: Yes

Reviewer #4: Yes

**********

6. Review Comments to the Author

Reviewer #1: The revision adequately addresses all the comments. I appreciate reading the nuances of the data. Well done.

Reviewer #3: PONE-D-20-36345R1

Racial and Ethnic Differentials in COVID-19-Related Job Exposures by Occupational Standing in the US

Thank you for the opportunity to review this revised manuscript. The authors present an interesting descriptive exercise trying to disentangle the occupational and ethnic/racial heterogeneity in exposure and risk of COVID-19. The document is well written and describes clearly the analytical process followed to arrive to their results.

Although the findings are very interesting and clearly presented, I consider that there are few methodological aspects that are key for the observed results and the interpretation, that are missing in this document. This is, confounding, the presence of heterogeneity and intersectionality. Given that the authors did not conduct any confounding adjustment, the robustness of the estimates is questionable. The authors did, however, stratify the analyses by the occupational standing and race/ethnicity, which highlights the issues with heterogeneity and the intersectionality (related to systemic racism, discriminations, opportunities for education, access to health care, and access and correct implementation of PPE, etc.) and call the attention to think about other several factors that could explain the results and that are not accounted for in this study. Therefore, I consider that as an initial approach to shed light into the issue of occupation and ethnicity in the disproportionate presence of COVID-19 outcomes among Black and Latinos, this is an adequate exercise. However, to properly address this issue, I would recommend a more robust approach, which would likely require other type of analysis and perhaps other type of data, which is beyond the scope of this manuscript. Hence, although I favour sharing this document with the academic community -again as a first step to identify the intricacies in the intersection between occupation and race/ethnicity in COVID-19- I would recommend some revisions.

Abstract:

In the body of the main manuscript the authors indicate:

Page 13; lines 268- 273 “Contrary to expectation based on earlier studies and the

media, the values in Fig 1 suggest that Black and Latino workers do not face higher occupational risks related to viral exposure compared with Whites. However, this conclusion is likely to be incorrect. The classification of high-risk occupations that we have considered thus far does not capture variation among occupations in COVID-19 risk reduction measures that are deployed in the workplace and are more likely to be implemented in higher status occupations.”

This aspect is key, the cumulative results (i.e, from the non-stratified analysis) are biased! Since the abstract is the door to the main manuscript, I recommend incorporating a sentence indicating that the overrepresentation of White in high-risk occupations is artifactual and incorrect, as you did in the results section of the manuscript. Likewise, I would recommend the consistent use of “lower (or higher) occupational standing” instead of “lower (or higher) occupational status”, to be consistent with the terminology used throughout the main text.

Introduction:

Line 74, please provide a reference for the statement and provide a short sentence about what has been considered essential vs frontline (you described “frontline” in page 8, but the “essential” definition is not provided).

Data:

Line 112: the ethics statement could be placed at the beginning or at the end of the section. In the current state, it cuts the flow for the information on the data sources.

Line 121: what are the six variables? I read the five types of “risk” but not the variables.

Race/ethnicity:

Do the authors consider that the “other Asian” category could be homogeneous? For instance, are Japanese and Pakistani similar groups? What could be the impact of merging these groups? How could this explain the results from Figure 4? Similar thoughts for “other race/ethnicity”. This could be mentioned in the discussion.

Line 155, please provide references for this statement.

Defining high risk occupations:

Lines 178-180. Which responses are you referring to? Also, could you indicate the implication of the mentioned correlation in the responses?

Analytic strategy:

Line 198-199: What could be the direction of the bias in presence of a differential response rate by ethnicity in the O*NET coupled with the differential labor participation rates? would suggest indicating this in the limitations.

I would recommend providing details of the analysis, including exclusion of observations, presence of missing data if any, software used for calculations and to generate the figures, etc.

Discussion

Line 371-373: using occupational standing for all the mentioned workers and workplace exposure to COVID-19 is an overstatement. There are several other aspects related to access, use and implementation of measures to decrease exposure to COVID-19 that are not solely based on occupational standing.

Limitations: I would recommend including the above-mentioned issues related to lack of adjustment for confounding, absence of evaluation of heterogeneity or interaction across variables and intersectionality, and the potential effects in the results and their interpretation. Likewise, please mention issues related to the misclassification of the work and therefore occupational standing?

Line 481-492: maybe goes beyond the scope of the manuscript and could be summarized.

Reviewer #4: Introduction

-The key “work-related factor” you are missing here is pay and access to healthcare. This means less ability to get care, less ability to take paid time off (increasing presenteeism), higher likelihood of living in crowded housing etc. Also for your point 1: Black, Latino and NA more likely to hold jobs that have to be done at their workplace rather than remotely…I am not sure this is totally true, it’s jobs that can’t be done from home which could mean they are working during the pandemic or could mean they have lost their job altogether (See Baker 2020 in AJPH: https://ajph.aphapublications.org/doi/full/10.2105/AJPH.2020.305738)

-Line 66: “we investigate these claims”—previously you cited peer-reviewed literature, seems reasonable to refer to that as more than a claim. Maybe “we investigate these hypotheses” or “these findings”. Claims makes it seems like you are telling occupational health researchers they are just making up things.

-Is there a reason you aren’t including median pay in your analysis? That is very easy to access and would give more weight to the “occupational standing”—and it’s easy to integrate with O*NET data

-Line 93: “lower status workers”—is this really want you mean? It’s only higher status because we as educated people have decided what is higher v. lower status work. Status really more an opinion than something that can be measured. Maybe “lower paid”, “more vulnerable”, “precarious” “underrepresented” ??

-Line 95: yes, underrepresented workers in more precarious jobs often have less access to risk mitigation strategies and workplace POWER than those in higher paid job. Worth mentioning the power element here. Maybe no one is getting proper PPE but there are some workers who could more easily ask than others

Data

-Can you defend why you used 1 year ACS data instead of 5 year? Especially for small demographic groups, 1 year is extremely noisy—5 year would be better. When you start stratifying by groups you start to lose reliability of the estimate with only 1 year ACS data.

See: https://www.census.gov/content/dam/Census/library/publications/2018/acs/acs_general_handbook_2018_ch03.pdf

-Worth noting O*NET isn’t just a sample of workers—also includes employers and job experts. (line 114)

-Line 120: virus is primarily transmitted through aerosols. This should at least be listed first (before respiratory droplets).

Results

-I’m quite confused about the numbers in your sample set. There are 160+ million workers in the USA. How come you only have so few? Given many of the groups that you report (Chinese, Filipino, Vietnamese, etc.) have less than 65,000 people, the 1 year ACS data is likely not very valuable and you should be using 5 year data instead to have more valid estimates.

-Line 251-253: this type of commentary seems more like discussion section fodder

-Your results assume that the (rudimentary) exposure metrics you included capture all occupational exposure risk with statements like “…Black and Latino workers do not face higher occupational risks related to viral exposure compared with Whites.” Instead what you mean to say is …”Black and Latino workers are less likely to be employed in occupations with characteristics related to increased exposure to SARS-CoV-2, as ascertained from five O*NET metrics.” Your clarifying sentence after this is also helpful and welcomed! However, do I think more of this should be in the discussion as opposed to in the results.

-Line 279: And be paid a living wage, have access to a union, etc.

Discussion

-I struggle (as another reviewer did) with saying that years of college education is a reasonable proxy for “occupational standing”. I think at a minimum wage should also be considered. This to me is a very “white collar” interpretation of occupational standing. But in the trades, where workers enjoy high pay, strong union protections, and access to benefits/safety…most have not completed a year of college. Pay/union representation seem equally important to me. Instead of using just the one variable as a proxy for “occupational standing” instead just refer to it as “educational attainment” with an understanding that we are imposing very white collar values on what makes “good work” and “important work”

-Line 375: “higher status”—suggest eliminating this phrasing as detailed in a previous comment

-Line 377: this gets at some of what I was detailing earlier, but I don’t think this necessarily relates to years of college. The trades/all goods producing occupations would fall into these categories even if their workers didn’t go to college (unless trade school is counted as college, but that seems unlikely)

-Line 383—another stray “status”

-Line 409: I get that there are differences between “essential” and “frontline” (very militaristic way of referring to these workers!) but wouldn’t the same limitations apply? “Frontline” would also have variability by states and over time? For example, if a restaurant is closed then those are not “frontline” workers, but if open then they would be, wouldn’t they?

Limitations

-Limitations with using 1 year ACS v. 5 year ACS especially for the small groups you are considering?

-Limitations because using the ACS means you have to combine some detailed occupations so you miss out on variability between occupations?

-Only able to look at a small subset not the whole workforce

-O*NET limitations: doesn’t cover all occupations

-O*NET limitations: subject to misclassification and undercounting, generated from subjective questionnaires, differential misclassification across groups (e.g. folks may not know they are exposed to disease/infection unless they are told they are and those risks are communicated making people in healthcare more likely to realize than other occupations), within-job exposure variation is not accounted for

-General limitations of your study: Many, many occupations/people are missing (some of which would be considered “frontline” and/or lower standing): independent contractors, domestic workers, military, is agriculture included?, self-employed, contingent, undocumented…These are also jobs that are more diverse

-General limitation: don’t include wages which is likely a larger predictor for how a pandemic would affect a worker

-Line 468: Disagree. If employers reduce workplace transmission, then community transmission is also reduced. We have to stop separating “occupational health” and “public health”. They are interconnected and if we protect workers, we are positively benefitting the community.

-Line 470: Thank you for giving some examples! However, these examples really only work for well paid, likely white workers. “preventing workers with symptoms or positive tests from coming to work” Does little for the Black taxi driver who has to make money in order to feed his family and taking a week off work could financially devastate him. How about instead, “providing paid sick leave that is ample and accessible, for which workers do not face retaliation or retribution for taking, if they have an exposure event” Or something like that.

-Line 475: is this still true about testing? I think we are at a point where we have a range of rapid and accurate tests…and isn’t all testing free even to the uninsured? People might not realize that, but it is.

-Line 486: Exposure controls are all the controls you are mentioning? Delete exposure. Also, “employers can provide PPE?” How about employers must or employers should

Figure 1

-Very blurry. Would be helpful to have the x-axis on the male section as well, it can be hard to follow it all the way down to the bottom

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #3: No

Reviewer #4: No

PLoS One. 2021 Sep 1;16(9):e0256085. doi: 10.1371/journal.pone.0256085.r004

Author response to Decision Letter 1

8 Jul 2021

We thank the reviewers for their helpful comments. Below we provide point-by-point responses to each of the reviewer’s comments. The reviewer’s comments are presented in italics (in the Word file).

Reviewer 1:

No further comments or corrections

Reviewer 3:

The objective of this analysis is to extend previous research which attempts to quantify work-related exposure to COVID-19 by race and ethnicity. Neither these previous studies nor our study attempts to provide a causal analysis to explain differences in risk across racial/ethnic groups by demographic and socioeconomic characteristics of the groups (e.g., by estimating multiple regression models) and so confounding is not an issue here. The goal of stratification by occupational standing is to provide a clearer picture of racial/ethnic differences in potential exposure to COVID-19 transmission. Occupational standing serves as a proxy (albeit an imperfect one) for workplace risk mitigation measures. We recognize that there are many variables that are important determinants of differential risk across racial and ethnic groups, such as those mentioned by the reviewer and those derived from factors beyond occupational exposure, but this type of analysis is not consistent with the objectives of this paper.

Page 13; lines 268- 273 “Contrary to expectation based on earlier studies and the

This issue is related to the previous concern. The non-stratified results are not biased nor is the overrepresentation of Whites in high-risk occupations artifactual or incorrect. These results portray the riskiness of occupations by race and ethnicity in the US population. A central point of our paper is precisely that this simple picture can be misleading because it ignores the fact that some race/ethnic groups in potentially high-risk jobs (e.g., physicians) are more likely to be protected at work than others. For example, Filipino frontline workers have a greater overrepresentation in occupations involving high risk of infection (such as healthcare work) than all other racial and ethnic groups. But, when we control for occupational standing, it becomes clear than many Filipinos in high risk environments are doctors and nurses. Our point is that although workers in higher occupational standing jobs may be employed in potentially hazardous environments (e.g., around many people with COVID-19), they are also much more likely than other workers to have protection, e.g., PPE, time and places to wash, highly frequent cleaning, etc., that reduces their risk of infection. Whether protection against COVID-19 transmission is available to workers is not captured in O*NET or other national data on workplace conditions. For this reason, we use occupational standing as a proxy for protection within the workplace. However, none of this negates the overall large overrepresentation of Filipinos in occupations with a high risk of infection, as derived from the O*NET measures.

We agree with the reviewer that the use of the term “occupational status” can be problematic. We have changed occupational status to occupational standing (OS) and, to avoid cumbersome text, we now use the terms low OS occupations and high OS occupations throughout the text in lieu of low status and high status occupations. It is important to keep in mind that the designation high or low standing does not refer to individual people, but rather to the standing of the occupation they hold compared to other occupations as based on the educational attainment of a national sample of workers in each occupation.

Introduction:

We have added a sentence with several citations that provide different definitions of essential occupations. There has been greater (although not complete) agreement on the definition of frontline than of essential.

Data:

Line 112: the ethics statement could be placed at the beginning or at the end of the section. In the current state, it cuts the flow for the information on the data sources.

We moved the ethics statement to the end of the section.

Line 121: what are the six variables? I read the five types of “risk” but not the variables.

We agree that this was confusing as written. We have revised several sentences in the Data section to clarify this. Two variables are used to capture working indoors. Risk 5 in the subsection “Defining high risk occupations” presents the two questions.

Race/ethnicity:

We definitely do not consider the “Other Asian” group to be homogenous. The ACS does not provide information regarding the groups that comprise the “Other race/ethnicity” category, but it is undoubtedly also a heterogeneous group. We have added “heterogenous” to the appropriate sentence in the text. Because of this heterogeneity and the relatively small size of these groups, we do not focus on them in the analysis.

Line 155, please provide references for this statement.

We added several references on this issue; we have placed them in the previous paragraph.

Defining high risk occupations:

Lines 178-180. Which responses are you referring to? Also, could you indicate the implication of the mentioned correlation in the responses?

We are referring to responses to the two questions on working outdoors (shown under risk 5). We have clarified this in the text. The strong correlation is a good sign since it suggests that both of these questions are capturing the same types of occupations.

Analytic strategy:

Unfortunately, O*NET does not provide information on the demographic characteristics of its survey respondents, so we have no idea whether non-response varies by race and ethnicity (or any other characteristics). The overall O*NET response rate is around 63% (O*NET, 2021) so it is possible that some groups were less likely to respond than others. Figuring out the consequences of potential differential underreporting in the absence of such information is complex. For example, if Latino workers were less likely to complete the O*NET survey and also more likely to be exposed to the general public or to work in close contact with coworkers than others in their same occupation, the potential risk score for that occupation would be artificially low. In contrast, if occupational incumbents included very few Latinos, the bias from Latino non-response would be minimal. Potential differential labor force participation rates add further complexity. While the reviewer raises a good point, it is impossible to provide meaningful estimates of the direction or magnitude of these types of biases. However, we have acknowledged the potential for non-response (from both the ACS and O*NET) in the text.

O*NET. 2021. Appendix E: Summary of Response Rate Experience to Date. https://www.onetcenter.org/dl_files/omb2021/AppendixE.pdf

I would recommend providing details of the analysis, including exclusion of observations, presence of missing data if any, software used for calculations and to generate the figures, etc.

We added information in the Analytic Strategy section on the number of persons in the ACS with military occupations (i.e., those excluded from the analysis because of the absence of O*NET information). No individuals in our sample of employed workers were missing information on education or race/ethnicity in the files provided by the US Census Bureau. We also added a sentence at the end of the Analytic Strategy indicating that STATA/MP Version 15.1 was used for calculations and figure production.

Discussion

As we say in the paper, we use occupational standing as a proxy for COVID-19 workplace mitigation efforts, not to represent occupational exposure nor as a determinant of workplace protections. As we discuss later in this section, we recognize that this is an imperfect measure that is subject to numerous limitations that we describe. Ideally, O*NET or another data source would have collected information on the implementation and regular use of workplace protections directly. In the absence of such data, we argue that occupational standing is a reasonable proxy for worker access to COVID-19 mitigation measures because extensive public health literature shows that “those at higher socioeconomic status… generally have access to a wider array of resources, including power and influence, to protect health than others do.” (pg. 5 in the manuscript). Specifically, we argue that “…higher [occupational standing] workers are more likely: (a) to work for employers who voluntarily practice risk mitigation and provide PPE and other tools for workers to do so, (b) if necessary, to demand risk reduction measures and to have the bargaining power to obtain them, and (c) to understand (or learn about) COVID-19 transmission routes and to comply with risk reduction strategies or implement their own.” In the Discussion section, we elaborate on the types of information that would have been desirable but are simply not available in O*NET – including the important factors that you mention.

We agree that adjustment for confounding, evaluation of heterogeneity and consideration of interactions would be important if this paper were intended as a causal or explanatory analysis. As we now indicate more explicitly at the beginning of the paper: “Our goal is to contribute insights into the pandemic by presenting the size and scope of these racial and ethnic disparities among workers rather than by estimating causal models of their determinants.” Because this is a descriptive analysis, confounding is not a problem, nor is the lack of interactions. In terms of intersectionality, the key innovation of our analysis (introducing OS as a proxy for risk mitigation) is a form of intersectional analysis: previous research did not provide as complete a picture as we do because we look at the intersection of race and ethnicity with OS (as well as gender). By considering the intersection of these variables, we develop a clearer picture of race and ethnic differentials in potential COVID-19 risk.

On the misclassification of work: respondents are not asked to pick an occupation from a list presented to them – many years of research have shown that this would lead to poor occupational reporting. Instead, respondents provide detailed verbal descriptions of their jobs and answer questions about their responsibilities and activities, the type of organization they work for or if they are self-employed, their job title, and the industry they work in. Trained occupational coders at the US Census Bureau use the verbal descriptions and other responses and a standardized system for occupational coding to code each respondent’s occupation. (The ACS questions (41 to 46) used to collect information for occupation coding are on page 11 at https://www2.census.gov/programs-surveys/acs/methodology/questionnaires/2018/quest18.pdf). Of course, even with these procedures, occupations can be misclassified, in the same way other data in surveys can be misreported by respondents. However, the use of detailed verbal work descriptions and standardized classification procedures using these verbal descriptions by trained classifiers reduces this potential. We now acknowledge in the Discussion section that misclassification of occupation could occur.

Line 481-492: maybe goes beyond the scope of the manuscript and could be summarized.

We added this section to the discussion at the explicit request of a PLOS ONE reviewer for the last round of reviews of this paper. We agree that it can be summarized and have done so.

Reviewer #4:

Introduction

Also for your point 1: Black, Latino and NA more likely to hold jobs that have to be done at their workplace rather than remotely…I am not sure this is totally true, it’s jobs that can’t be done from home which could mean they are working during the pandemic or could mean they have lost their job altogether (See Baker 2020 in AJPH: https://ajph.aphapublications.org/doi/full/10.2105/AJPH.2020.305738)

As we try to make clearer in the introduction, the goal of the paper is to describe potential workplace risks of exposure to COVID-19 by race and ethnicity – in order to assess whether workplace conditions are likely to play an important role in the higher rates of COVID-19 infection among Black, Latino, Native American, and some other groups of workers. We agree that access to healthcare and the ability to take time off from work, as well as living in crowded households, living in neighborhoods with more infected or fewer vaccinated people, etc., affect the probability that workers bring SARS-CoV-2 into their workplace and therefore, expose co-workers to the virus. However, accounting for the multiple reasons that workers in disadvantaged groups may be more likely to contract COVID-19 is beyond the scope of this analysis.

As we noted above, we use occupational standing in the analysis, not as a measure of an individual’s socioeconomic status, but as a proxy variable for the likelihood that workers in each occupation have access to risk mitigation measures in their workplace. We describe the logic of doing so both in the paper and in response to Reviewer #3’s comments. Occupational standing is based on the average value of those employed in the occupation rather than on an individual measure – thus, it is an occupation-level, not an individual-level, variable. We have revised the section in the paper on “Occupational Standing” to explain why we chose occupational education over occupation income:

We chose occupational education rather than occupational income – another well-established indicator of occupational ranking – both because educational attainment is typically reported with greater accuracy than income or wages and because among those with a valid occupational code in the ACS data file provided by IPUMS, educational attainment information is complete while information on income is missing for one-fifth of individuals. The correlation between these two measures of occupational ranking across the occupations in the dataset is high (0.72).

As we also note in the paper, our analytic procedure does not permit inclusion of those who lost their job, a loss that varies by race and ethnicity. To further emphasize this issue, we have added a sentence emphasizing your point above (in the Analytic Strategy section) and citing the Baker paper.

This is a matter of word usage in different disciplines. We have changed ‘claims” to “issues” in the Abstract and we have rewritten the relevant sentence in the Introduction.

See our comment above with regard to why we prefer occupational education to occupational income.

Occupational status or standing is a sociological term representing objective location (as defined by a specific variable or set of variables) in the social hierarchy. The field of sociology has an extensive literature on how to measure occupational status or standing consistently and reliably. As we noted in response to Reviewer #3, who had similar concerns, we no longer use terms such as lower status or higher status workers and occupations. Instead, we refer to occupational standing, or OS for clarity, which is defined clearly in the text, in terms of an average educational level for all workers in that occupation.

We agree and have modified the text. We also note in the Discussion section that workers in high OS occupations are more likely “to demand risk reduction measures and have the bargaining power to obtain them” (1st paragraph) and later in the Discussion section that workers in low OS occupations have “less control and decision-making ability over how their workplaces are run.”

Data

See: https://www.census.gov/content/dam/Census/library/publications/2018/acs/acs_general_handbook_2018_ch03.pdf

Yes, we can defend this decision. We now note in the first sentence of the Data section that occupation is based on a response to a question on the current or most recent job or business held in the five years prior to the ACS. We also note (in the Analytic Strategy section) that 88% of the responses to this question pertain to jobs or businesses held in the one year prior to the ACS. If we were to use the 5-year ACS, many of the jobs in our analysis would not be very recent ones (e.g., they could go back as far as a decade, which would include, for example, the Great Recession), making this information less relevant to the period of the pandemic. We emphasize in the Discussion section the importance of capturing jobs close to the pandemic period.

The ACS samples a large number of people each year. In our data set we have about 1.9 million respondents who reported jobs. As you can see from Table 1, the smallest groups have about 3,000 recent workers each and only three groups have below 10,000 recent workers; there are over 160,000 Black recent workers and over 250,000 Latino recent workers. These sample sizes are much larger than those in almost every other survey and sufficiently large to stratify by quartiles of occupational standing.

-Worth noting O*NET isn’t just a sample of workers—also includes employers and job experts. (line 114)

Yes, you are correct. We have moved the reference to workers to the subsequent sentence to clarify that the data we use on hazardous job characteristics come from incumbent workers.

-Line 120: virus is primarily transmitted through aerosols. This should at least be listed first (before respiratory droplets).

We have reversed the order of aerosols and respiratory droplets.

Results

Altogether, our sample includes nearly 2 million workers. As we noted earlier, for this analysis we chose to prioritize the recency of the data, leading to our use of the 1-year ACS. The guidance from the Census Bureau that you reference regarding the use of the 5-year ACS for populations below 65,000 refers to geographic areas rather than racial/ethnic subgroups.

See: https://www.census.gov/programs-surveys/acs/guidance/estimates.html

-Line 251-253: this type of commentary seems more like discussion section fodder

We reworded this sentence.

Your results assume that the (rudimentary) exposure metrics you included capture all occupational exposure risk with statements like “…Black and Latino workers do not face higher occupational risks related to viral exposure compared with Whites.” Instead what you mean to say is …”Black and Latino workers are less likely to be employed in occupations with characteristics related to increased exposure to SARS-CoV-2, as ascertained from five O*NET metrics.” Your clarifying sentence after this is also helpful and welcomed! However, do I think more of this should be in the discussion as opposed to in the results.

We toned down the sentence regarding Black and Latino workers to make the result more tentative.

-Line 279: And be paid a living wage, have access to a union, etc.

We agree that there are many other key differences between physicians and home health workers, including pay, union access and others. However, our point here was that workers in two occupations with a high level of potential exposure to the coronavirus through contact with infected or potentially infected people may differ considerably in access to PPE and other risk mitigation measures.

Discussion

As we note above, occupational education is not the same as individual educational attainment. Instead, it is an average of the educational attainment (percent with some college education) of all workers in a particular occupation. Occupational education and occupational income are two standard sociological indicators of location within the social hierarchy – standing which empirically provides access to power and control. We agree that union representation can be very important, but we don’t have any information on this issue.

See our earlier response as to why we use occupational education rather than occupational income. Use of both variables would be problematic since they are highly correlated. We include this information in the revised text.

-Line 375: “higher status”—suggest eliminating this phrasing as detailed in a previous comment; Line 383—another stray “status”

Done.

Again, keep in mind that occupational education refers to an average educational attainment for all workers in a particular occupation, not for an individual. We agree that for some subgroups of occupations, the occupational standing variable (the percent of workers within each occupation who have completed at least a year of college) may not do as good a job as other variables at predicting which workers have more or less access to COVID-19 mitigation measures. Obviously, it would be ideal to have a direct measure of use of mitigation measures in each workplace. Since such a direct measure is not available and we are comparing across a very wide range of occupations, we believe that occupational standing, defined here as occupational education, does a reasonable job as a proxy for occupations in which workers are more likely to have the power and control necessary to insure COVID-19 mitigation measures in their workplace. As noted above, occupational education is highly correlated with occupational income and so substituting occupational income as the proxy would not greatly change our results.

We have adopted the language (essential and frontline) used in studies of work exposure during the pandemic. Our point is that the definition of “essential” workers – which was defined by federal and state agencies – has changed frequently over the course of the pandemic and across space. In contrast, the definition of “frontline” we use is static – it is defined based on an assessment (using O*NET variables) of whether or not jobs could be done at home. It seems much less likely that jobs would change rapidly or frequently from being able to be done at home to not being so. You are correct that there is variability in terms of which workers are frontline primarily because many frontline workers lost their jobs due to closure (although some reductions in frontline exposure may be temporary). We now emphasize this issue of job loss in the second paragraph of the Analytic Strategy. We also mention the issue of job loss when we describe limitations of our analysis in the Discussion section. Nevertheless, frontline worker status is more straightforward to define and has resulted in a more consistent classification across studies than essential worker status.

Limitations

-Limitations with using 1 year ACS v. 5 year ACS especially for the small groups you are considering?

-Limitations because using the ACS means you have to combine some detailed occupations so you miss out on variability between occupations?

-Only able to look at a small subset not the whole workforce

-O*NET limitations: doesn’t cover all occupations

We explained above why use of the 1-year ACS is preferable to the 5-year ACS. The ACS includes a very large high-quality unbiased random sample of the entire workforce. Moreover, O*NET does cover almost all occupations: we are able to link almost all occupations with O*NET (an exception is the military; we have added information on the number of ACS respondents in the military). In response to your concern, we added a sentence in the Discussion regarding the limitation of having to combine detailed occupations that may reflect different risk exposures.

We have added sentences regarding response error in the O*NET surveys (as well as the ACS). At the end of the paragraph, we address your concern regarding within-job exposure variation (we note such variation by age, sex, and race/ethnicity of workers).

ACS uses a population-based sample that includes respondents regardless of type of work or other characteristics such as documentation status. Particular types of jobs are missing from our analysis only if none of the nearly 2 million respondents in the ACS reported holding that job (in the previous five years). In these cases, the jobs were simply not held during this period rather than being technically “missing.” The sample includes independent contractors, domestic workers, military personnel, agricultural workers, self-employed workers and those in contingent employment. Although there may be higher non-response rates among undocumented workers, ACS data have been used to estimate the size and characteristics of the undocumented population specifically because adults are included regardless of documentation status (Capps et al., 2018). As noted above, with the exception of military workers who are not included in O*NET, we are able to match virtually all occupations held by ACS workers with O*NET job characteristics. We agree that there is diversity in activities and exposure within jobs, but there is not much we can do about it.

Note that the occupations listed in S2 Appendix include only the most frequently held occupations in each OS quartile (and by risk factor). Thus, many occupations we include in our analysis are not listed in this appendix.

Capps R, Bachmeier JD, Van Hook J. Estimating the Characteristics of Unauthorized Immigrants Using U.S. Census Data: Combined Sample Multiple Imputation. The ANNALS of the American Academy of Political and Social Science. 2018;677: 165–179. doi:10.1177/0002716218767383

-General limitation: don’t include wages which is likely a larger predictor for how a pandemic would affect a worker

As we describe above, we are not attempting to predict how the pandemic would affect individual workers.

We completely agree that workplace transmission is a part of community transmission and have rewritten that sentence in the Discussion section to acknowledge this interdependence.

We have included your example of sick leave in our revised discussion.

We have added a sentence in the Introduction to indicate that our results best reflect the situation at the end of 2020, although they continue to be relevant.

-Line 486: Exposure controls are all the controls you are mentioning? Delete exposure. Also, “employers can provide PPE?” How about employers must or employers should

In the previous version, we describe examples of the full range of controls outlined by Carlsten et al. (citation in the paper references) based on industrial hygiene’s hierarchy of controls model. We did so at the explicit request of a PLOS ONE reviewer for the last round of reviews of this paper. At the request of Reviewer #3 for this round of the paper, we have now summarized and shortened this section.

Figure 1

-Very blurry. Would be helpful to have the x-axis on the male section as well, it can be hard to follow it all the way down to the bottom

We suspect the Figure became less clear when we ran it through the required software. We have now created a higher resolution figure and added the racial/ethnic groups to the male graph.

Attachment

Submitted filename: Response to Reviewers.docx

Click here for additional data file.^{(43.5KB, docx)}

PLoS One. doi: 10.1371/journal.pone.0256085.r005

Decision Letter 2

Marlene Camacho-Rivera

22 Jul 2021

PONE-D-20-36345R2

Racial and Ethnic DIfferentials in COVID-19-Related Job Exposures by Occupational Standing in the US

PLOS ONE

Dear Dr. Goldman,

==============================

ACADEMIC EDITOR:

Please note that I have requested 2 minor but important edits to acknowledge the limitations of the dataset (with respect to workers that may be excluded or misclassified) and the descriptive nature of the study (to temper interpretations around causality). Once these edits have been made, I will accept the manuscript. No other changes are required.

Please submit your revised manuscript within 30 days of receipt of this email. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.
A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.
An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

We look forward to receiving your revised manuscript.

Kind regards,

Marlene Camacho-Rivera, ScD, MPH

Academic Editor

PLOS ONE

Journal Requirements:

Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

Additional Editor Comments (if provided):

I appreciate the considerable effort that has gone into revising this manuscript twice and have forwarded this decision without the additional review. There are 2 very minor edits, which can be resolved quickly, but are important given the potential contributions of the paper:

1. Please make more explicit the limitations of the O*NET sampling frame and mention the types of workers that are left out of the O*NET sampling frame.

2. Please make more explicit the descriptive nature of the study, both in the study objectives and limitations section.

With these 2 edits, the manuscript will be accepted.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

Reviewer #3: All comments have been addressed

Reviewer #4: (No Response)

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

Reviewer #3: Partly

Reviewer #4: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #3: I Don't Know

Reviewer #4: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

Reviewer #3: Yes

Reviewer #4: No

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

Reviewer #3: Yes

Reviewer #4: Yes

**********

6. Review Comments to the Author

Reviewer #3: The authors have addressed the majority of the comments/aspects raised by the reviewers. However, even if both reviewers raised concerns about a limitations section, the authors did not incorporate it. Although, authors included some sentences throughout the discussion about the concerns mentioned and the manuscript has been greatly improved, there are some issues that remain and are worthy of clarification. For instance, the issues of confounding and heterogeneity raised still hold. For instance, the response indicates that the study is only descriptive. However, this has not been stated anywhere in the text and the objective of "extending previous work on racial/ethnic differences in potential work-related exposure to COVID-19 by examining 12 racial/ethnic groups and by considering five indicators of potential risk exposure" including to "examine potential risk of exposure separately by occupational standing (OS)," may not be understood as only descriptive. Given the use of "control for OS", as stated by the authors in their answer. If the authors did not intended the estimation a causal effect and association, nor the interpretation as such, this could be explicitly presented in the abstract, objective/methods and discussion section of the manuscript as well, so the interpretations could be accordingly made.

Reviewer #4: Thanks to the authors for the improvements to the paper, both minor and major.

I have just a few lingering concerns after which adopted, I will consider the paper to be publication ready.

A few places in your comments to reviewer you emphasize that O*NET covers all occupations, and that the ACS survey captures all occupations. I think the fallacy you are falling into is the assumption that the ~700 SOC codes covers all workers in the United States which is just not the case (you do acknowledge that military is not included in O*NET). Given the way O*NET is collected (where they reach out to workers/employers/experts about each of the occupations covered by SOCs), many workers are left out of the data collection—for example, Uber drivers wouldn’t be surveyed by O*NET, or a delivery worker for Amazon, a temporary worker for any organization, or a domestic worker being paid under the table, or even a general contractor, or a migrant agricultural worker. These workers may be surveyed by ACS, but their reported occupation is being matched to a census occupation code that isn’t very precise, and then you are further matching to O*NET data that was generated to represent a different population of workers (i.e. only those in more traditional employer-employee relationships covered by a SOC code). Employee respondents to O*NET are found through their employer. The authors acknowledge there is inner-occupation variability, but “there is not much we can do about it.” I am only asking for you to acknowledge these limitations—(1)collapsing all the diverse occupations self-reported in the ACS data to 400 occupations induces inner-occupation variability, and (2) in matching ACS to O*NET data, you are comparing two different populations (ACS is general US cross section, O*NET is a very specific set of occupations marked by a typical employee-employer relationship and is only available in English and Spanish) that have different distributions of occupational variability.

Neither of these limitations negate your work, but it has to be mentioned that O*NET does not cover all workers and indeed many of the workers it does not cover may be more “frontline” or at increased risk (and also could be more likely to be workers of color). By not acknowledging these “missing” workers that are continuously left out of BLS data collection it is further perpetuating their decreased standing and lack of importance and necessity to our economy.

Moreover, the lack of data from O*NET on the demographics of respondents makes it hard to ensure the sample is representative of the American workforce, though most assume that it is.

I would suggest reviewing this paper if you haven’t already: https://onlinelibrary.wiley.com/doi/full/10.1002/ajim.20846?casa_token=MnvFp1NvrwUAAAAA%3ACffACaImtyoCu67O96UX_SBe49mfnSxZWlzrI8Rzkr9xmbeuL0uv3kVWI16Xb8geBNN2JnpXV_9n1Jg

Thank you for helping me understand the use of 1 year v. 5 year ACS. I agree that in this analysis the currency of the data is preferable to the precision of the data.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #3: No

Reviewer #4: No

PLoS One. 2021 Sep 1;16(9):e0256085. doi: 10.1371/journal.pone.0256085.r006

Author response to Decision Letter 2

26 Jul 2021

We thank the reviewers for their helpful comments. Below we provide responses to each of the reviewer’s comments. The reviewer’s comments are presented in italics.

Reviewer 3:

The authors have addressed the majority of the comments/aspects raised by the reviewers. However, even if both reviewers raised concerns about a limitations section, the authors did not incorporate it. Although, authors included some sentences throughout the discussion about the concerns mentioned and the manuscript has been greatly improved, there are some issues that remain and are worthy of clarification. For instance, the issues of confounding and heterogeneity raised still hold. For instance, the response indicates that the study is only descriptive. However, this has not been stated anywhere in the text and the objective of "extending previous work on racial/ethnic differences in potential work-related exposure to COVID-19 by examining 12 racial/ethnic groups and by considering five indicators of potential risk exposure" including to "examine potential risk of exposure separately by occupational standing (OS)," may not be understood as only descriptive. Given the use of "control for OS", as stated by the authors in their answer. If the authors did not intended the estimation a causal effect and association, nor the interpretation as such, this could be explicitly presented in the abstract, objective/methods and discussion section of the manuscript as well, so the interpretations could be accordingly made.

As noted previously, we did not intend this paper to provide a causal analysis. As suggested by the reviewer, we now explicitly state that the analysis is descriptive. We have added wording to the Abstract, Introduction, Analytic Strategy and Discussion sections to make this very clear. For example, in the Discussion we now say the following: “The analysis is descriptive – portraying potential workplace risks by race, ethnicity and gender – rather than an investigation of the root causes of these patterns.”

Reviewer #4:

Thanks to the authors for the improvements to the paper, both minor and major.

I have just a few lingering concerns after which adopted, I will consider the paper to be publication ready.

Moreover, the lack of data from O*NET on the demographics of respondents makes it hard to ensure the sample is representative of the American workforce, though most assume that it is.

Thank you for helping me understand the use of 1 year v. 5 year ACS. I agree that in this analysis the currency of the data is preferable to the precision of the data.

We have changed the text in several places in the Discussion to make these points. We acknowledge that the ACS classification scheme has resulted in the combining of O*NET occupations thereby increasing intra-occupation variability. In a subsequent paragraph, we discuss the likely consequence of O*NET surveys being restricted to formal workers – i.e., underestimation of the extent of racial and ethnic differences in workplace exposures to COVID-19. We acknowledge that we are probably capturing substantial numbers of informal workers in the ACS data.

Attachment

Submitted filename: Response to Reviewers.docx

Click here for additional data file.^{(22.1KB, docx)}

PLoS One. doi: 10.1371/journal.pone.0256085.r007

Decision Letter 3

Marlene Camacho-Rivera

30 Jul 2021

Racial and Ethnic DIfferentials in COVID-19-Related Job Exposures by Occupational Standing in the US

PONE-D-20-36345R3

Dear Dr. Goldman

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Marlene Camacho-Rivera, ScD, MPH

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

Reviewer #4: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

Reviewer #4: (No Response)

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #4: (No Response)

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

Reviewer #4: (No Response)

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

Reviewer #4: (No Response)

**********

6. Review Comments to the Author

Reviewer #4: (No Response)

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #4: No

PLoS One. doi: 10.1371/journal.pone.0256085.r008

Acceptance letter

Marlene Camacho-Rivera

6 Aug 2021

PONE-D-20-36345R3

Racial and ethnic differentials in COVID-19-related job exposures by occupational standing in the US

Dear Dr. Goldman:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Marlene Camacho-Rivera

Academic Editor

PLOS ONE

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

S1 Appendix. Racial/ethnic distributions within each occupational subset.

(DOCX)

Click here for additional data file.^{(68.4KB, docx)}

S2 Appendix. Common high-risk occupations by occupational standing quartile.

(DOCX)

Click here for additional data file.^{(34.6KB, docx)}

Attachment

Submitted filename: Response to Reviewers.docx

Click here for additional data file.^{(23.3KB, docx)}

Attachment

Submitted filename: Response to Reviewers.docx

Click here for additional data file.^{(43.5KB, docx)}

Attachment

Submitted filename: Response to Reviewers.docx

Click here for additional data file.^{(22.1KB, docx)}

Data Availability Statement

[pone.0256085.ref001] 1.Andrasfay T, Goldman N. Reductions in 2020 US life expectancy due to COVID-19 and the disproportionate impact on the Black and Latino populations. Proc Natl Acad Sci. 2021;118. doi: 10.1073/pnas.2014746118 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0256085.ref002] 2.Andrasfay T, Goldman N. Association of the COVID-19 Pandemic With Estimated Life Expectancy by Race/Ethnicity in the United States, 2020. JAMA Netw Open. 2021;4: e2114520–e2114520. doi: 10.1001/jamanetworkopen.2021.14520 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0256085.ref003] 3.Leggat-Barr K, Uchikoshi F, Goldman N. COVID-19 Risk Factors and Mortality among Native Americans. medRxiv. 2021; 2021.03.13.21253515. doi: 10.1101/2021.03.13.21253515 [DOI] [Google Scholar]

[pone.0256085.ref004] 4.Rentsch CT, Kidwai-Khan F, Tate JP, Park LS, Jr JTK, Skanderson M, et al. Patterns of COVID-19 testing and mortality by race and ethnicity among United States veterans: A nationwide cohort study. PLOS Med. 2020;17: e1003379. doi: 10.1371/journal.pmed.1003379 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0256085.ref005] 5.Sáenz R, Garcia MA. The Disproportionate Impact of COVID-19 on Older Latino Mortality: The Rapidly Diminishing Latino Paradox. J Gerontol Ser B. 2020;XX: 1–7. doi: 10.1093/geronb/gbaa158 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0256085.ref006] 6.Wang D, Gee GC, Bahiru E, Yang EH, Hsu JJ. Asian-Americans and Pacific Islanders in COVID-19: Emerging Disparities Amid Discrimination. J Gen Intern Med. 2020. [cited 6 Nov 2020]. doi: 10.1007/s11606-020-06264-5 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0256085.ref007] 7.Woolf SH, Masters RK, Aron LY. Effect of the covid-19 pandemic in 2020 on life expectancy across populations in the USA and other high income countries: simulations of provisional mortality data. BMJ. 2021;373: n1343. doi: 10.1136/bmj.n1343 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0256085.ref008] 8.Casura L, Lowe R Jr., Martinez C, Serpas S, Castellanos V, Singelmann J. Frontline workers in the U.S.: race, ethnicity, and gender. International Union for the Scientific Study of Population; 2020Sep. Available: https://www.niussp.org/article/frontline-workers-in-the-u-s-race-ethnicity/ [Google Scholar]

[pone.0256085.ref009] 9.Hawkins D. Differential occupational risk for COVID-19 and other infection exposure according to race and ethnicity. Am J Ind Med. 2020;65. doi: 10.1002/ajim.23145 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0256085.ref010] 10.Covid-19 Deaths Skew Younger Among Minorities—WSJ. The Wall Street Journal. 17 Aug 2020. Available: https://www-wsj-com.libproxy1.usc.edu/articles/covid-19-deaths-strike-early-for-many-minorities-11597582801. Accessed 10 Nov 2020.

[pone.0256085.ref011] 11.Oppel Jr RA, Gebeloff R, Lai KKR, Wright W, Smith M. The Fullest Look Yet at the Racial Inequity of Coronavirus. The New York Times. 5 Jul 2020. Available: https://www.nytimes.com/interactive/2020/07/05/us/coronavirus-latinos-african-americans-cdc-data.html. Accessed 10 Nov 2020.

[pone.0256085.ref012] 12.Reckdahl K. “The impact of work”: On-the-job coronavirus exposure a key driver in Black, Latino communities. NOLA.com. 17 Oct 2020. Available: https://www.nola.com/news/business/article_d0e20ff2-0a62-11eb-8314-af3dffbffd15.html. Accessed 10 Nov 2020.

[pone.0256085.ref013] 13.U.S. News and World Report. AHA News: Hispanics Are Paying the Price for Being “Essential” During the Pandemic. US News & World Report. 9 Oct 2020. Available: https://www.usnews.com/news/health-news/articles/2020-10-09/aha-news-hispanics-are-paying-the-price-for-being-essential-during-the-pandemic. Accessed 10 Nov 2020.

[pone.0256085.ref014] 14.Erdman SL. More Hispanic workers impacted by Covid-19 in food processing and agriculture workplaces, CDC study finds. CNN. 20 Oct 2020. Available: https://www.cnn.com/2020/10/20/health/cdc-study-covid-hispanic-workers/index.html. Accessed 10 Nov 2020.

[pone.0256085.ref015] 15.Cox-Ganser JM, Henneberger PK. Occupations by Proximity and Indoor/Outdoor Work: Relevance to COVID-19 in All Workers and Black/Hispanic Workers. Am J Prev Med. 2021. [cited 30 Mar 2021]. doi: 10.1016/j.amepre.2020.12.016 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0256085.ref016] 16.Dubay L, Aarons J, Brown KS, Kenney GM. How Risk of Exposure to the Coronavirus at Work Varies by Race and Ethnicity and How to Protect the Health and Well-Being of Workers and Their Families. Urban Institute; 2020Dec. Available: https://www.urban.org/research/publication/how-risk-exposure-coronavirus-work-varies-race-and-ethnicity-and-how-protect-health-and-well-being-workers-and-their-families/view/full_report [Google Scholar]

[pone.0256085.ref017] 17.Carlsten C, Gulati M, Hines S, Rose C, Scott K, Tarlo SM, et al. COVID-19 as an occupational disease. Am J Ind Med. 2021;64: 227–237. 10.1002/ajim.23222 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0256085.ref018] 18.Hawkins D, Davis L, Kriebel D. COVID-19 deaths by occupation, Massachusetts, March 1–July 31, 2020. Am J Ind Med. 2021;64: 238–244. doi: 10.1002/ajim.23227 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0256085.ref019] 19.Bailey ZD, Krieger N, Agénor M, Graves J, Linos N, Bassett MT. Structural racism and health inequities in the USA: evidence and interventions. The Lancet. 2017;389: 1453–1463. doi: 10.1016/S0140-6736(17)30569-X [DOI] [PubMed] [Google Scholar]

[pone.0256085.ref020] 20.del Río C, Alonso-Villar O. The Evolution of Occupational Segregation in the United States, 1940–2010: Gains and Losses of Gender–Race/Ethnicity Groups. Demography. 2015;52: 967–988. doi: 10.1007/s13524-015-0390-5 [DOI] [PubMed] [Google Scholar]

[pone.0256085.ref021] 21.Chetty R, Hendren N, Jones MR, Porter SR. Race and Economic Opportunity in the United States: an Intergenerational Perspective*. Q J Econ. 2020;135: 711–783. doi: 10.1093/qje/qjz042 [DOI] [Google Scholar]

[pone.0256085.ref022] 22.Pager D, Shepherd H. The Sociology of Discrimination: Racial Discrimination in Employment, Housing, Credit, and Consumer Markets. Annu Rev Sociol. 2008;34: 181–209. doi: 10.1146/annurev.soc.33.040406.131740 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0256085.ref023] 23.Selden TM, Berdahl TA. COVID-19 And Racial/Ethnic Disparities In Health Risk, Employment, And Household Composition. Health Aff (Millwood). 2020;39: 1624–1632. doi: 10.1377/hlthaff.2020.00897 [DOI] [PubMed] [Google Scholar]

[pone.0256085.ref024] 24.Blau FD, Koebe J, Meyerhofer PA. Who are the Essential and Frontline Workers? National Bureau of Economic Research; 2020 Sep. Report No.: w27791. doi: 10.3386/w27791 [DOI] [PMC free article] [PubMed]

[pone.0256085.ref025] 25.Chen Y-H, Glymour M, Riley A, Balmes J, Duchowny K, Harrison R, et al. Excess mortality associated with the COVID-19 pandemic among Californians 18–65 years of age, by occupational sector and occupation: March through October 2020. medRxiv. 2021; 2021.01.21.21250266. doi: 10.1101/2021.01.21.21250266 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0256085.ref026] 26.Phelan JC, Link BG, Tehranifar P. Social Conditions as Fundamental Causes of Health Inequalities: Theory, Evidence, and Policy Implications. J Health Soc Behav. 2010;51: S28–S40. doi: 10.1177/0022146510383498 [DOI] [PubMed] [Google Scholar]

[pone.0256085.ref027] 27.Gaitens J, Condon M, Fernandes E, McDiarmid M. COVID-19 and Essential Workers: A Narrative Review of Health Outcomes and Moral Injury. Int J Environ Res Public Health. 2021;18: 1446. doi: 10.3390/ijerph18041446 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0256085.ref028] 28.Billock RM, Groenewold MR, Free H, Haring Sweeney M, Luckhaupt SE. Required and Voluntary Occupational Use of Hazard Controls for COVID-19 Prevention in Non–Health Care Workplaces—United States, June 2020. MMWR Morb Mortal Wkly Rep. 2021;70. doi: 10.15585/mmwr.mm7007a5 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0256085.ref029] 29.Hoven H, Wahrendorf M, Siegrist J. Occupational position, work stress and depressive symptoms: a pathway analysis of longitudinal SHARE data. J Epidemiol Community Health. 2015;69: 447–452. doi: 10.1136/jech-2014-205206 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0256085.ref030] 30.Landsbergis PA, Grzywacz JG, LaMontagne AD. Work organization, job insecurity, and occupational health disparities. Am J Ind Med. 2014;57: 495–515. doi: 10.1002/ajim.22126 [DOI] [PubMed] [Google Scholar]

[pone.0256085.ref031] 31.Clougherty JE, Souza K, Cullen MR. Work and its role in shaping the social gradient in health. Ann N Y Acad Sci. 2010;1186: 102–124. doi: 10.1111/j.1749-6632.2009.05338.x [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0256085.ref032] 32.Ndugga N, Pham O, Hill L, Artiga S, Parker N. Latest Data on COVID-19 Vaccinations by Race/Ethnicity. San Francisco, CA: Kaiser Family Foundation; 2021Jun. Available: https://www.kff.org/coronavirus-covid-19/issue-brief/latest-data-on-covid-19-vaccinations-race-ethnicity/ [Google Scholar]

[pone.0256085.ref033] 33.Ruggles S, Flood S, Goeken R, Grover J, Meyer E, Pacas J, et al. IPUMS USA: Version 10.0 [dataset]. Minneapolis, MN: IPUMS; 2020. Available: 10.18128/D010.V10.0 [DOI] [Google Scholar]

[pone.0256085.ref034] 34.O*NET OnLine. National Center for O*NET Development; Available: https://www.onetonline.org/

[pone.0256085.ref035] 35.Wilson N, Corbett S, Tovey E. Airborne transmission of covid-19. BMJ. 2020;370. doi: 10.1136/bmj.m3206 [DOI] [PubMed] [Google Scholar]

[pone.0256085.ref036] 36.Hauser RM, Warren JR. Socioeconomic Indexes for Occupations: A Review, Update, and Critique: Sociol Methodol. 2016. [cited 6 Nov 2020]. Available: https://journals.sagepub.com/doi/10.1111/1467-9531.271028 [Google Scholar]

[pone.0256085.ref037] 37.Tomer A, Kane JW. How to protect essential workers during COVID-19. Washington, D.C.: The Brookings Institution; 2020Mar. Available: https://www.brookings.edu/research/how-to-protect-essential-workers-during-covid-19/ [Google Scholar]

[pone.0256085.ref038] 38.Rogers TN, Rogers CR, VanSant-Webb E, Gu LY, Yan B, Qeadan F. Racial Disparities in COVID-19 Mortality among Essential Workers in the United States. World Med Health Policy. 2020; 10.1002/wmh3.358. doi: 10.1002/wmh3.358 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0256085.ref039] 39.Dingel JI, Neiman B. How Many Jobs Can be Done at Home? National Bureau of Economic Research; 2020 Apr. Report No.: w26948. doi: 10.3386/w26948 [DOI]

[pone.0256085.ref040] 40.Unemployment rates by age, sex, race, and Hispanic or Latino ethnicity. Washington, DC: U.S. Bureau of Labor Statistics; 2020 Oct. Report No.: E-16.

[pone.0256085.ref041] 41.Gemelas J, Davison J, Keltner C, Ing S. Inequities in Employment by Race, Ethnicity, and Sector During COVID-19. J Racial Ethn Health Disparities. 2021. [cited 30 Mar 2021]. doi: 10.1007/s40615-021-00963-3 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0256085.ref042] 42.Baker MG. Nonrelocatable Occupations at Increased Risk During Pandemics: United States, 2018. Am J Public Health. 2020;110: 1126–1132. doi: 10.2105/AJPH.2020.305738 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0256085.ref043] 43.StataCorp. Stata Statistical Software: Release 15. College Station, TX:StataCorp LLC.: StataCrop; 2017. [Google Scholar]

[pone.0256085.ref044] 44.Sterling MR, Tseng E, Poon A, Cho J, Avgar AC, Kern LM, et al. Experiences of Home Health Care Workers in New York City During the Coronavirus Disease 2019 Pandemic: A Qualitative Analysis. JAMA Intern Med. 2020;180: 1453–1459. doi: 10.1001/jamainternmed.2020.3930 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0256085.ref045] 45.Shang J, Chastain AM, Perera UGE, Dick AW, Fu CJ, Madigan EA, et al. The state of infection prevention and control at home health agencies in the United States prior to COVID-19: A cross-sectional study. Int J Nurs Stud. 2021;115: 103841. doi: 10.1016/j.ijnurstu.2020.103841 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0256085.ref046] 46.Choy CC. Empire of Care: Nursing and Migration in Filipino American History. Duke University Press; 2003. [Google Scholar]

[pone.0256085.ref047] 47.Hall M, Greenman E. The Occupational Cost of Being Illegal in the United States: Legal Status, Job Hazards, and Compensating Differentials. Int Migr Rev. 2015;49: 406–442. doi: 10.1111/imre.12090 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0256085.ref048] 48.Wilson FD. Generational changes in racial inequality in occupational attainment, 1950–2010: A synthetic cohort analysis. Bois Rev. 2017;14: 387–425. doi: 10.1017/S1742058X17000170 [DOI] [Google Scholar]

[pone.0256085.ref049] 49.Steege AL, Baron SL, Marsh SM, Menéndez CC, Myers JR. Examining occupational health and safety disparities using national data: A cause for continuing concern. Am J Ind Med. 2014;57: 527–538. doi: 10.1002/ajim.22297 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0256085.ref050] 50.Krieger N, Chen JT, Waterman PD, Hartman C, Stoddard AM, Quinn MM, et al. The inverse hazard law: Blood pressure, sexual harassment, racial discrimination, workplace abuse and occupational exposures in US low-income black, white and Latino workers. Soc Sci Med. 2008;67: 1970–1981. doi: 10.1016/j.socscimed.2008.09.039 [DOI] [PubMed] [Google Scholar]

[pone.0256085.ref051] 51.Quandt SA, LaMonto NJ, Mora DC, Talton JW, Laurienti PJ, Arcury TA. COVID-19 Pandemic among Latinx Farmworker and Nonfarmworker Families in North Carolina: Knowledge, Risk Perceptions, and Preventive Behaviors. Int J Environ Res Public Health. 2020;17: 5786. doi: 10.3390/ijerph17165786 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0256085.ref052] 52.McClure ES, Vasudevan P, Bailey Z, Patel S, Robinson WR. Racial Capitalism Within Public Health—How Occupational Settings Drive COVID-19 Disparities. Am J Epidemiol. 2020;189: 1244–1253. doi: 10.1093/aje/kwaa126 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0256085.ref053] 53.U.S. Bureau of Labor Statistics. Unemployment Rate [UNRATE], retrieved from FRED. Federal Reserve Bank of St. Louis; 2020 Nov. Available: https://fred.stlouisfed.org/series/UNRATE

[pone.0256085.ref054] 54.U.S. Bureau of Labor Statistics. Unemployment Rate—Hispanic of Latino [LNS14000009], retrieved from FRED. Federal Reserve Bank of St. Louis; 2020 Nov. Available: https://fred.stlouisfed.org/series/LNS14000009

[pone.0256085.ref055] 55.U.S. Bureau of Labor Statistics. Unemployment Rate—Black or African American [LNS14000006], retrieved from FRED. Federal Reserve Bank of St. Louis; 2020 Nov. Available: https://fred.stlouisfed.org/series/LNS14000006

[pone.0256085.ref056] 56.Kochhar R. Fewer mothers and fathers in U.S. are working due to COVID-19 downturn; those at work have cut hours. Washington, D.C.: Pew Research Center; 2020Oct. Available: https://www.pewresearch.org/fact-tank/2020/10/22/fewer-mothers-and-fathers-in-u-s-are-working-due-to-covid-19-downturn-those-at-work-have-cut-hours/ [Google Scholar]

[pone.0256085.ref057] 57.Protecting Workers: Guidance on Mitigating and Preventing the Spread of COVID-19 in the Workplace | Occupational Safety and Health Administration. Washington, DC: Occupational Safety & Health Administration; 2021Jan. Available: https://www.osha.gov/coronavirus/safework [Google Scholar]

[pone.0256085.ref058] 58.Michaels D, Wagner GR. Occupational Safety and Health Administration (OSHA) and Worker Safety During the COVID-19 Pandemic. JAMA. 2020;324: 1389–1390. doi: 10.1001/jama.2020.16343 [DOI] [PubMed] [Google Scholar]

PERMALINK

Racial and ethnic differentials in COVID-19-related job exposures by occupational standing in the US

Noreen Goldman

Anne R Pebley

Keunbok Lee

Theresa Andrasfay

Boriana Pratt

Roles

Abstract

Introduction

Data

Variables

Race/ethnicity

Occupational standing

Frontline status

Defining high risk occupations

Analytic strategy

Results

Table 1. Number of recent workers and percentage in frontline occupations by race/ethnicity and sex.

Fig 1. Representation of frontline workers in high-risk occupations relative to representation as frontline workers, by race/ethnicity and sex.

Fig 2. Representation of frontline workers in high-risk occupations relative to representation as frontline workers by occupational standing: White, Black and Latino workers, by sex.

Fig 3. Representation of frontline workers in high-risk occupations relative to representation as frontline workers by occupational standing: Chinese, Filipino, and Vietnamese workers, by sex.

Fig 5. Representation of frontline workers in high-risk occupations relative to representation as frontline workers by occupational standing: Pacific Islander, mixed race, and other race workers, by sex.

Fig 4. Representation of frontline workers in high-risk occupations relative to representation as frontline workers by occupational standing: Korean, other Asian, and native American workers, by sex.

Discussion

Supporting information

Acknowledgments

Data Availability

Funding Statement

References

Decision Letter 0

Marlene Camacho-Rivera

Roles

Author response to Decision Letter 0

Decision Letter 1

Marlene Camacho-Rivera

Roles

Author response to Decision Letter 1

Decision Letter 2

Marlene Camacho-Rivera

Roles

Author response to Decision Letter 2

Decision Letter 3

Marlene Camacho-Rivera

Roles

Acceptance letter

Marlene Camacho-Rivera

Roles

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases