SARS-CoV-2 Seroprevalence among Healthcare, First Response, and Public Safety Personnel, Detroit Metropolitan Area, Michigan, USA, May–June 2020

To estimate seroprevalence of severe acute respiratory syndrome 2 (SARS-CoV-2) among healthcare, first response, and public safety personnel, antibody testing was conducted in emergency medical service agencies and 27 hospitals in the Detroit, Michigan, USA, metropolitan area during May–June 2020. Of 16,403 participants, 6.9% had SARS-CoV-2 antibodies. In adjusted analyses, seropositivity was associated with exposure to SARS-CoV-2–positive household members (adjusted odds ratio [aOR] 6.18, 95% CI 4.81–7.93) and working within 15 km of Detroit (aOR 5.60, 95% CI 3.98–7.89). Nurse assistants (aOR 1.88, 95% CI 1.24–2.83) and nurses (aOR 1.52, 95% CI 1.18–1.95) had higher likelihood of seropositivity than physicians. Working in a hospital emergency department increased the likelihood of seropositivity (aOR 1.16, 95% CI 1.002–1.35). Consistently using N95 respirators (aOR 0.83, 95% CI 0.72–0.95) and surgical facemasks (aOR 0.86, 95% CI 0.75–0.98) decreased the likelihood of seropositivity.

coordinate an emergency medical services (EMS) system for a geographic region, to invite employees to participate. The protocol was reviewed by CDC human subjects research officials, who determined the activity to be public health surveillance and exempt from full institutional review board review (2).

Study Participants
Eligible participants for the serology survey included adults >18 years of age who worked onsite in a first response, hospital, or public safety setting and consented to phlebotomy and serum sample storage for confirmation of test results if needed. Persons were not eligible to participate if, in the 2 weeks before taking the survey, they reported having new symptoms of cough, shortness of breath, or change in sense of taste or smell, or had tested positive for SARS-CoV-2 by RT-PCR test using a nasal, throat, or saliva sample.

Web-Based Survey
Participating agencies shared information about the secure web-based survey (Appendix Table 1, https:// wwwnc.cdc.gov/EID/article/26/12/20-3764-App1. pdf) with employees using email and onsite marketing. There was no face-to-face recruitment. Participation was voluntary and individual results were not shared with employers. CDC did not have access to personal identifiers. The survey was drafted by the investigators, reviewed by CDC subject matter experts, and was designed to require <10 minutes to complete on a personal device. Upon survey completion, participants received information about blood collection sites at their workplace or a nearby MCA location.

Statistical Analysis
Of 16,403 participants, 6 (0.04%) had samples that were unable to be tested and were excluded (n = 16,397). Percent SARS-CoV-2 antibody positivity and 95% CIs were calculated. Statistical testing was conducted using Cochran-Armitage trend tests for variables with ordinal categories (2-sided tests with α = 0.05). Non-Hispanic Native Hawaiian and other Pacific Islander (n = 31, 0.2%), Non-Hispanic American Indian/Alaska Native (n = 53, 0.3%), and other race (n = 320, 2.0%) participants were categorized as other race. The 398 (2.4%) participants who declined to report race were categorized separately and included in all analyses.
Participants could choose multiple work locations; 17.5% chose >1 location. Each work location category was represented as a separate dichotomous variable (i.e., dummy variable) to enable modeling of non-mutually exclusive categories. Participants were provided with occupation categories and a free text option. The National Institute for Occupational Safety and Health (NIOSH) assisted with coding free-text responses using the NIOSH Industry and Occupation Computerized Coding System (4). No categories created from free-text options reached high sample size (n<100) and were coded as "other" except for technicians (e.g., dialysis, telemetry, surgery), which were combined into a "clinical technician" category (n = 365).
Exposure to persons with confirmed COVID-19 (co-worker, household member, patient, and other person) was defined as contact within 6 feet for >10 minutes, but the question did not mention PPE use, which was assessed in separate questions. PPE use was dichotomized for each piece of equipment into "use all the time" (the recommended, or optimal, frequency when PPE is required) versus all other choices. Similarly, exposure to persons with confirmed COVID-19 was dichotomized into "yes" versus all other choices.
Differences between categories were assessed by nonoverlapping 95% CIs for percent positivity. Two participants were excluded from adjusted analyses (1 participant with missing housing information and 1 participant from a nonstudy hospital; n = 16,395). To account for clustering of participants by facility/agency, generalized estimating equations were used to model the likelihood of seropositivity. Covariates were chosen a priori to represent risk of exposure and infection. Model diagnostics performed with regression analysis did not show evidence of collinearity for work location (highest values for variance inflation factor = 1.4, and condition index = 4.1), which was represented by non-mutually exclusive dummy variables entered simultaneously into the multivariable model. No interaction terms were explored. SAS 9.4 software (https://www.sas.com) was used for all analyses. ArcGIS (ESRI, https://www.esri.com) was used to map seroprevalence by agency location.
By work location, seroprevalence was highest among participants who worked in hospital wards (8.8%, 95% CI 8.0%-9.7%) and lowest among those working in police departments (3.9%, 95% CI 2.5%-5.8%) ( Table 2). Within hospitals, lower seroprevalence was found among persons working in intensive care units (ICUs; 6.1%) and operating rooms or surgical units (4.5%) compared with participants working in wards (8.8%) and emergency departments (EDs; 8.1%). By occupation, the highest †Test for trend in seropositivity statistically significant at α = 0.05. ‡Other race/ethnicity includes non-Hispanic Native Hawaiian and other Pacific Islander, non-Hispanic American Indian and Alaska Native, and participants who indicated other race.
§Categories are not mutually exclusive.
Other occupational risk factors are included in the Appendix. Participants reported the average number of times per shift since March 1, 2020, in which they had participated in aerosol-generating procedures (Appendix Figure 2) and were given a list of examples for reference (Appendix Table 1). Seroprevalence generally increased with increasing procedure frequency (p = 0.04 by test for trend), with the highest percent positivity among those who participated in such procedures >25 times per shift on average (9.1%, 95% CI 7.4%-11.0%). Participants also reported how frequently they used each component of PPE, using a Likert scale. Overall, there was no pattern seen in percent antibody positivity with frequency of use with any PPE component (Appendix Table 2). Among those reporting ideal frequency of use ("all the time") for a specific PPE component, seroprevalence was similar to the overall seroprevalence (Table 2).
Multivariable adjustment using generalized estimating equations was performed (Figure 3; Appendix Table 3). Factors most strongly associated with likelihood of seropositivity were exposure to a household member with confirmed COVID- 19 (8). Our study revealed ≈2.5 times more infections than cases based on selfreported RT-PCR results. The 2.7% positivity for RT-PCR may be higher in the healthcare and first responder population compared with the general public (which ranged from 0.22% to 1.04% in the 6 Michigan counties) as a result of targeted and repeated testing of personnel in hospitals and emergency medical services settings (11). Even so, surveillance of these occupational groups in Detroit based on self-reported RT-PCR testing results would have identified a minority of infections. Healthcare workers are known to be at occupational risk for SARS-CoV-2 exposure (12). Participants in occupations that may involve frequent and prolonged patient contact, such as nurse assistants and nurses (13,14), were more likely to be seropositive than physicians. Multivariable analysis revealed a weak association between lower seropositivity and consistent use of N95 respirators and surgical facemasks. Lower seroprevalence was observed among participants who reported high use of PPE despite shortages and reuse/extension protocols that could be hypothesized to lower the observed effectiveness of PPE. These and other confounding factors may obscure the role PPE plays in preventing infection, and it may be necessary to account for multiple factors in studies assessing the effect of PPE. The lower likelihood of seroprevalence associated with working in the controlled environments of a hospital ICU or surgical ward may reflect the impact of additional mitigation measures, including clear identification of infected persons and environmental and engineering controls (15). This pattern of lower seropositivity among staff in higher-risk versus lower-risk hospital settings has been described previously (16). However, even within healthcare work settings, some workers such as nurse assistants had a higher risk of infection than those in other roles. This finding highlights the concern that certain occupations may require additional focus on assessing and controlling factors related to transmission.
Together, these analyses of community and workplace factors show the contribution of community acquired infection to seropositivity among Detroit area healthcare workers. For 3 hospital settings (hospital ward, ED, and ICU) that could be compared across healthcare occupations, seropositivity rose with closer proximity of the facility to the Detroit center. This pattern suggests that regardless of occupation or work location, community acquisition was a common underlying factor of infection risk. There are 2 related implications. First, the observed impact of PPE may be reduced given the background impact of community acquisition of SARS-CoV-2 infection. Second, reducing community spread through population-based measures may directly protect healthcare workers on 2 fronts: reduced occupational exposure as a result of fewer infected patients in the less controlled workplace settings such as the ED, and reduced exposure in their homes and communities.
After adjusting for other factors, we found that women were less likely than men to be seropositive. This pattern was seen only in adjusted analysis; women's lower risk may have been obscured by their disproportionate representation in the occupations at higher risk of infection. Women represented 69% of the sample but made up 86% of those in nursing and nurse assistance.
Participants >65 years of age were less likely to be infected than younger workers. This pattern may be the result of measures to protect older workers from high-risk situations or from greater precautions taken among this group. A population study that also observed lower seroconversion among older persons found that older persons were less likely to live with a household contact (17). Seroconversion may also diminish with age in general (17), although other studies showed no pattern by age or higher seroprevalence among older persons (18,19). Participants of non-Hispanic Black race/ethnicity remained more likely to be seropositive than non-Hispanic white participants, even after adjustment. Community-level surveillance of COVID-19 infection and SARS-CoV-2 infection has demonstrated overrepresentation of minority groups in population-adjusted analyses (20,21). One hypothesis for the higher risk of infection among Black and Hispanic persons is employment in jobs without possibility of working remotely (22). Unfortunately, the survey did not collect information about occupation and workplace of household members. We speculate that the higher risk of exposure/infection among non-Hispanic Black versus non-Hispanic White participants in our study likely reflects uncontrolled confounding by factors for which data were not available. Some limitations must be considered. The survey was a convenience sample with unknown representativeness: 80% of the 20,650 employees anticipated by MCA and hospital contacts to be eligible participated but agency participation varied, with highest participation among hospital personnel. The crosssectional design precluded determining the source of exposure. In addition, comprehensive exposure data (e.g., travel, commuting, social exposures) were not collected. Because of the limited questionnaire length, PPE questions did not probe donning and doffing training, participant familiarity with PPE use, or reuse or extension protocols that may have affected effectiveness (11). No additional questions were asked about other workplace infection control practices. Another potential source of bias is the healthy worker effect, in which persons with prolonged COVID-19 infection or sequelae would not have been onsite to participate. Seroprevalence may be underestimated, given that the sensitivity of the antibody test was less than 100%. It is also possible that participants who were infected did not seroconvert (23; F. Gallais et al., unpub. data, https://www. medrxiv.org/content/10.1101/2020.06.21.2013244 9v1), but it is unknown whether lack of seroconversion may have occurred systematically between occupations (e.g., those exposed more intensely or with more severe illness may be more likely to develop antibodies) (24). Although more recent infections may have not been detected, it is unlikely that this varied systematically across groups. Strengths included coverage of a large number of personnel at hospitals and first response/public safety facilities and pairing antibody testing with questionnaire data to enable focus on a high-risk population.
Key implications for the risk of SARS-CoV-2 infection among healthcare, first response, and public safety personnel include the impact of community acquisition, increased odds of exposure associated with specific healthcare occupations, and the protection provided by PPE. Effects of interventions that could be further studied and implemented include providing alternative housing to healthcare workers during times or in areas of high community prevalence and ensuring that workers in high-risk occupations are given adequate PPE, specifically N95 respirators and surgical facemasks, as well as infection control training.