Measuring work-related risk of COVID-19: comparison of COVID-19 incidence by occupation and industry – Wisconsin, September 2020-May 2021

Abstract Background Work-related exposures play an important role in SARS-CoV-2 transmission, yet few studies have measured the risk of COVID-19 across occupations and industries. Methods During September 2020 – May 2021, the Wisconsin Department of Health Services collected occupation and industry data as part of routine COVID-19 case investigations. Adults aged 18-64 years with confirmed or probable COVID-19 in Wisconsin were assigned standardized occupation and industry codes. Cumulative incidence rates were weighted for non-response and calculated using full-time equivalent (FTE) workforce denominators from the 2020 American Community Survey. Results An estimated 11.6% of workers (347,013 of 2.98 million) in Wisconsin, ages 18-64 years, had COVID-19 from September 2020 to May 2021. The highest incidence by occupation (per 100 full-time equivalents) occurred among personal care and services workers (22.4), healthcare practitioners and support staff (20.7), and protective services workers (20.7). High risk sub-groups included nursing assistants and personal care aides (28.8), childcare workers (25.8), food and beverage service workers (25.3), personal appearance workers (24.4), and law enforcement workers (24.1). By industry, incidence was highest in healthcare (18.6); the highest risk sub-sectors were nursing care facilities (30.5) and warehousing (28.5). Conclusions This analysis represents one of the most complete examinations to date of COVID-19 incidence by occupation and industry. Our approach demonstrates the value of standardized occupational data collection by public health, and may be a model for improved occupational surveillance elsewhere. Workers at higher risk of SARS-CoV-2 exposure may benefit from targeted workplace COVID-19 vaccination and mitigation efforts.

Occupations requiring close contact with customers and co-workers have been linked to workforce 3 shortages [3], severe disease [4] and death [5] among workers due to COVID-19. . While many 4 epidemiologic studies on occupational COVID-19 risk have focused on healthcare workers [6][7][8][9][10][11][12], 5 the risks of COVID-19 are present in a wide variety of work settings [2]. This has been 6 demonstrated by outbreaks at manufacturing and food processing facilities [13,14], correctional 7 facilities [15], and other high-density work settings [16][17][18] throughout the pandemic. 8 Despite the importance of occupation in determining one's risk of SARS-CoV-2 exposure, 9 relatively few studies have compared COVID-19 risk across occupation and industries in the United 10 States. Prior studies have compared hospitalizations or deaths by occupation [4,5,19], or the 11 frequency of outbreaks by industry [20, 21], but have not been able to assess individual exposure 12 risk across different work settings. This gap is due, in part, to a lack of standardization in the 13 collection and reporting of occupational data among U.S. public health systems. Poor occupational 14 data for COVID-19 has not only led to delays in identification and response to workplace outbreaks, 15 but has limited our ability to identify occupations and industries that are at high-risk for SARS-CoV-2 16 transmission and target these workers with public health resources and policy considerations [22]. 17 To address this gap, in June 2020, CDC recommended that U.S. public health jurisdictions begin 18 collecting detailed occupation and industry information for all COVID-19 cases in a standardized 19 format to facilitate occupational coding and surveillance [22]. This approach was implemented by the 20 Wisconsin Department of Health Services (WDHS) in September 2020. 21 This report utilizes the first eight months of Wisconsin's standardized occupational data 22 collection (September 2020-May 2021) to calculate COVID-19 incidence by occupation and industry. 23 Our observation period coincides with the first major COVID-19 surge in Wisconsin, prior to 24 widespread COVID-19 vaccination, and after Wisconsin's "Safer At Home" order had expired (May 25 2020), which brought many workers back to in-person jobs. As one of the first U.S. jurisdictions to 26 A C C E P T E D M A N U S C R I P T

Incidence among major industry sectors 14
The highest cumulative incidence and greatest number of COVID-19 cases occurred in the 15 Healthcare industry (NAICS 62; n = 71,531), with an incidence of 18.6 per 100 FTE (Fig 3)

Incidence among industry sub-sectors 21
Nursing and residential care facilities had the highest incidence (30.5 per 100 FTE) among 22 all industry sub-sectors included in this analysis (Appendix 1, Figure S4). Warehousing and storage 23 facilities (NAICS 493) ranked second among industry sub-sectors with an incidence of 28.5 per 100 24 We estimated the incidence of COVID-19 by occupation and industry in Wisconsin during 7 September 2020 -May 2021. Overall, 11.6% of Wisconsin workers had confirmed or probable 8 COVID-19 during the observation period (12.3 per 100 FTE), representing a high risk of COVID-19 9 to workers during this time. 10 Personal Care and Service occupations, a group that includes childcare workers, 11 hairdressers, and other services jobs, experienced the highest incidence of COVID-19 (22.1 per 100 12 FTE) in our analysis. These jobs often require close contact with clients and may involve exposure to 13 SARS-CoV-2 without the same level of institutional controls available in healthcare settings. High 14 incidence among personal appearance workers (hair stylists, manicurists, etc.) was consistent with 15 their high-risk designation (close proximity, indoor, public-facing) in the SARS-CoV-2 Occupational 16 Exposure Matrix (SOEM) [30], as well as studies showing poor ventilation in salon settings [31]. 17 Childcare workers, the broad occupation with the highest incidence in this group, provided essential 18 in-person services during this period. High incidence among these workers highlights the risks 19 experienced in this setting where masking and social distancing might have been challenging. 20 Healthcare practitioners and support staff experienced the second highest incidence in our 21 analysis (20.7 per 100 FTE). This is consistent with multiple prior studies showing high incidence in 22 this group [6][7][8][9][10][11][12] . The highest risk sub-group in our analysis were support staff comprising of 23 nursing assistants, home health aides, and personal care assistants. Prior studies have also found 24 high incidence in this group [6,32]. This sub-group is commonly employed in nursing care facilities, 25 a sub-sector that has experienced frequent outbreaks [33], and, in our study, had the highest 26 incidence among all industry sub-sectors. Within nursing care facilities, health care support workers 27 were disproportionately affected, representing 38% of workers in these facilities but nearly half 1 (48%) of all COVID-19 cases in the residential care sub-sector (others included food staff, 2 healthcare providers, maintenance workers, and managers). Nursing assistants in nursing care 3 facilities are also more likely to hold second jobs compared to other healthcare workers, increasing 4 the potential for outbreaks to cross workplaces [34]. 5 The high incidence of COVID-19 found among Protective Service occupations (20.7 per 100 6 FTE; 3 rd highest occupational group) in Wisconsin was also observed among law enforcement and 7 first responders in an Arizona cohort [35], and is consistent with their designation in SOEM as high-8 risk due to frequent close contact with the public [30]. Two other U.S. seroprevalence studies early in 9 2020, however, did not find elevated risk in this group [6,36]. The longer timespan of our study, 10 which occurred prior to widespread vaccination and during a period of substantial transmission in 11 Wisconsin may account for this difference. The fact that Wisconsin correctional facilities experienced 12 several large COVID-19 outbreaks in fall 2020 [15] likely contributed to high incidence in this group, 13 and to correctional officers having the second highest incidence among all broad occupations in 14 Wisconsin. 15 Workers in Food Service and Retail Trade experienced high COVID-19 incidence during the 16 observation period. These workers are likely to have prolonged exposure to unmasked persons, and 17 are less likely than other occupations to have access to paid leave [37], exacerbating workplace 18 risks for this group. Within this sub-group, bartenders experienced the highest risk (37.0 per 100 19 FTE), and the highest risk among all broad occupations. This is consistent with a Norwegian study 20 that identified bartenders as the occupation with the highest incidence after pandemic lockdowns 21 were lifted [38]. 22 With respect to industry, high-risk sectors largely aligned with analogous high-risk 23 occupations (i.e., healthcare, food service, public safety) discussed above. One exception was 24 warehouse facilities, which had the second highest incidence among all industry sub-sectors. This 25 sector experienced frequent outbreaks during 2020-2021 [20,33], and the large number of materials 26 handlers, transportation workers, and production workers on-site could explain observed risk 27 estimates. Another notable industry sub-sector was food manufacturing, which had a lower 1 incidence than expected (13.8 per 100 FTE; 16 th ranked sub-sector). Outbreaks in this sector were 2 widely reported in Wisconsin in spring 2020 [13], prior to data collection for this study. Thus, many 3 workers had recovered from recent infections, before for the observation period, which could have 4 led to underestimation of risk in this high-density workplace. 5

Strengths 6
There are several notable strengths of our approach. First, this work represents one of the 7 largest and most complete examinations to date of COVID-19 risk among occupations and 8 industries. This led to identification of high incidence rates among several previously under-9 recognized groups such as personal appearance workers, childcare workers, food service workers, 10 and others. Second, our integration of NIOCCS auto-coded industry and occupation information into 11 routine COVID-19 case interviews is novel. NIOCCS has become an important tool for analyzing 12 occupational risk factors for a variety of diseases, but has primarily been used retrospectively [39, 13 40]. Our real-time data capture and coding represents a strong model for occupational surveillance 14 that could benefit other U.S. jurisdictions. Third, our study benefitted from the opportune timing of the 15 observation period during September 2020 to May 2021. This period was characterized by high 16 incidence in Wisconsin, widespread availability of COVID-19 testing, and participation in case 17 investigation interviews (75% of confirmed and probable cases were reached for interview during 18 this period). This time period was also after the Wisconsin "Safer At Home" order was lifted in May 19 2020, when many workers had returned to in-person work. Emergence of variants and proliferation 20 of at-home antigen tests later in 2021 led to declines in case reporting, follow-up, and interview 21 completion in Wisconsin. This likely increased representativeness and reduced the impact of 22 reporting or testing biases in our analysis. 23

Limitations 24
These findings are subject to several limitations. First, it was not possible to distinguish between 25 exposures that occurred at the workplace versus other locations (e.g., community, household) in this 26 analysis. Thus, risk estimates for each occupation or industry could be affected by social or 27 behavioral risk factors unrelated the specific work setting if such factors are differentially distributed 1 across occupations and industries. Second, 2020 ACS estimates for workforce size are considered 2 experimental. Certain groups, particularly low-income and racial and ethnic minority groups, may be 3 underrepresented in ways that could affect occupational estimates [41]. Third, despite efforts to 4 supplement case interview data with other available data sources, industry and occupation inputs 5 were missing for 43% and 47% of eligible cases for this analysis, respectively. The use of non-6 response weights to account for missing data, while powerful, were likely not able to account for all 7 sector-specific differences in response probability. Lastly, our adjustment methods could not 8 account for differences in testing behaviors between occupations and industries. Mandatory 9 screening testing in some industries or increased availability of workplace or community testing 10 options could have biased reported estimates. 11 12

Conclusions 13
In this analysis, we described COVID-19 incidence by occupation and industry in Wisconsin. 14 Our findings highlighted the high incidence of COVID-19 in Wisconsin among workers in service 15 occupations and the healthcare industry during September 2020 -May 2021, and identified multiple 16 occupational sub-groups that were particularly impacted during this peak period of transmission. 17 Groups at increased risk of workplace exposure to SARS-CoV-2 could benefit from continued efforts 18 to promote COVID-19 vaccination, booster coverage, and other setting-specific mitigation strategies 19 such as mask use, symptom screening, improved ventilation, and testing when indicated by local 20 conditions. 21 More broadly, collection of occupational data for COVID-19 cases in many U.S. states 22 remains limited to outbreaks, specific jobs-of-interest, or other non-standardized data formats. Occupation Computerized Coding System (NIOCCS) and provided invaluable technical assistance 7 throughout this project; thank you to the hundreds of COVID-19 case interviewers at the state and 8 local health departments in Wisconsin who contributed data to this report. 9

Disclaimer 10
The findings and conclusions of this report are those of the author(s) and do not necessarily 11 represent the official position of the Centers for Disease Control and Prevention.  1 ¶ The number of cases reported represents the final weighted estimates for case totals in each category after non-2 response adjustment, after excluding cases among all non-paid or unemployed persons (e.g., retired, student, 3 volunteer, homemaker) and the armed forces.

†
The reference value used for risk ratio calculations among major occupation and industry groups was the 5 combined incidence across all groups.

‡
Other race categories represented among cases ("Native Hawaiian or Pacific Islander", "Multiple Races", 7 "Unknown" and "Other") were not able to be calculated due to non-concordance with race categories given in 8 ACS denominator data.