Risk Factors Associated With SARS-CoV-2 Seropositivity Among US Health Care Personnel

Key Points Question What risk factors are associated with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) seropositivity among health care personnel (HCP) inside and outside the workplace? Findings In this cross-sectional study of 24 749 HCP in 3 US states, contact with an individual with known coronavirus disease 2019 (COVID-19) exposure outside the workplace was the strongest risk factor associated with SARS-CoV-2 seropositivity, along with living in a zip code with higher COVID-19 incidence. None of the assessed workplace factors were associated with seropositivity. Meaning In this study, most risk factors associated with SARS-CoV-2 infection among HCP were outside the workplace, suggesting that current infection prevention strategies in health care are effective in preventing patient-to-HCP transmission in the workplace.


Introduction
2][3][4][5][6] It remains unclear whether certain job functions or specific workplace activities, including care for individuals with known and unknown SARS-CoV-2 positivity, increase the risk of SARS-CoV-2 infection.Furthermore, it is still unknown whether the association of individual and job-related characteristics and risk of SARS-CoV-2 infection are consistent across health care systems and over time.
To address these knowledge gaps, 4 sites in the US Centers for Disease Control and Prevention (CDC) Prevention Epicenters Program, based in academic institutions that collaborate with each other and the CDC to perform innovative infection prevention research, used previously collected serosurvey data to determine the prevalence of antibodies to SARS-CoV-2 in a large multistate study of US health care personnel (HCP).We sought to identify risk factors associated with seropositivity, including HCP demographic characteristics, work location, work exposure to patients with COVID-19, and community exposure to COVID-19 with the a priori hypothesis that community exposure but not health care exposure was associated with seropositivity.Our study builds on a prior single Prevention Epicenter cross-sectional study 7 that explored specific workplace activities unique to that system.

Participating Academic Centers
We assessed SARS-CoV-2 seroprevalence in large health care systems affiliated with 4 Prevention Epicenters in Atlanta, Georgia (Emory Healthcare), Baltimore, Maryland (Johns Hopkins Medicine and University of Maryland Medical System), and Chicago, Illinois (Rush University System).The systems were primarily based in metropolitan areas and predominantly acute care hospitals, but they included more than 100 affiliated regional ambulatory locations, administrative locations, rehabilitation facilities, and skilled nursing facilities.
This study was reviewed by each site's institutional review board and either approved or determined to be non-human participant research.At each site where this activity was deemed human participant research, a waiver for informed consent was obtained or individual informed consent was obtained.This activity was reviewed by the CDC and was conducted consistent with applicable federal law and CDC policy (eg, 45 CFR part 46; 21 CFR part 56; 42 USC §241(d); 5 USC §552a; 44 USC §3501).This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Serologic Testing, Data Collection, and Data Compilation
All badged HCP at each site were eligible to participate.Each site independently designed and conducted a voluntary HCP serological survey.The activities were started and implemented as part of internal quality and occupational safety assessments or research activities in each health care system.At the time of specimen collection, HCP completed a site-specific survey, including occupational activities and possible exposures to individuals with SARS-CoV-2 infection both inside and outside the workplace.Race and ethnicity were self-reported by HCP using categories defined by

JAMA Network Open | Infectious Diseases
Risk Factors Associated With SARS-CoV-2 Seropositivity Among US Health Care Personnel each site and were included in this study because of prior reports of racial disparities in SARS-CoV-2 seropositivity. 4,8,9Subsequently, we combined specimen and survey data collected between the weeks of April 19 and August 30, 2020.
The serological test used at each of the sites met the US Food and Drug Administration emergency use criteria and all measured immunoglobin G. 10 The Abbott Architect assay (Abbott Laboratories), targeting the nucleocapsid protein, was used at Rush.Emory used a laboratorydeveloped assay, 11 and Johns Hopkins used the QuantiVac ELISA (EUROIMMUN), both of which target the receptor binding domain of the SARS-CoV-2 spike protein. 12For the University of Maryland, only samples testing positive by both the VITROS assay (Ortho Clinical Diagnostics) and a laboratory-developed ELISA, both targeting the spike protein, were interpreted as positive for this study.
Sites shared a limited data set (without protected health information) for the combined analysis, including date and results of serology, 3-digit prefix of participants' residential zip code, and survey responses.Because survey content varied between sites, questions from each survey were mapped to a common metadata set across the 4 sites.In some instances, text or ordinal values were mapped to categorical values to allow inclusion of data from all facilities, while in other instances substantial differences in the surveys prohibited definitive mapping.Mapping of key health care exposure variables are detailed in eTable 1 in the Supplement.Variables that could not be mapped definitively across facilities due to differences in question wording or response categories were excluded.The date of specimen collection was converted to the week number since January 1, 2020.
Age was categorized by decade (ie, <30, 30-39, 40-49, 50-59, and Ն60 years) in the limited data set.Workplace location was categorized as emergency department, inpatient (regardless of direct care of patients with COVD-19), other locations (ambulatory, perioperative, surgical, rehabilitation or postacute care, no patient contact, worked from home), or unknown location.Because working in COVID-19-focused areas was captured differently across all sites, COVID-19 care was categorized as never providing COVID-19 care or providing any COVID-19 care.HCP were categorized into mutually exclusive job roles.
In addition to self-reported community contact with individuals with COVID-19, we assigned a zip code-based value of community exposure to COVID-19 to each HCP.Weekly data on the number of COVID-19 cases by 3-digit zip code prefix were obtained from the state departments of health in Georgia, Illinois, and Maryland.These data were combined with census population data to calculate weekly COVID-19 cumulative incidence by 3-digit zip code.Each participant in our metadata set was then assigned the value of COVID-19 cumulative incidence in their zip code of residence until 1 week prior to their test date.

Statistical Analysis
We calculated the prevalence of SARS-CoV-2 seropositivity overall (combined across all sites) and for each site.The overall seropositivity was first modeled using unadjusted bivariate logistic regression assessing the association of each individual potential risk factor with seropositivity.We developed a mixed-effects logistic regression model, with a random intercept to account for clustering by the 4 sites, to estimate adjusted odds ratios (aORs) and 95% CIs between potential risk factors and SARS-CoV-2 seropositivity across all 4 sites.Model variables included demographic characteristics (sex, age group, race, ethnicity), community contact, COVID-19 cumulative incidence (base-10 logarithm of the COVID-19 cumulative incidence), and occupational factors (job role, workplace location, any contact with a patient with COVID-19).Three sensitivity analyses were performed.We chose to include a random intercept in our main model to produce effect estimates representing the association between workplace factors and seropositivity across all 4 health care systems combined.A sensitivity analysis using logistic regression without a random intercept was performed and compared with the main model to determine whether including the random intercept improved model fit.A second sensitivity analysis included a time variable (month of test) in the model to account for monthly variations in

JAMA Network Open | Infectious Diseases
Risk Factors Associated With SARS-CoV-2 Seropositivity Among US Health Care Personnel unmeasured confounders.Lastly, we applied our main model to a subset of data excluding Emory to assess whether key risk factors identified in a detailed Emory-only analysis 7 remained risk factors in this larger analysis.Analyses were conducted in R version 4.0.2(R Project for Statistical Computing) using the lme4 package, and the code is publicly available. 13No prespecified level of statistical significance was set.

Setting and Participants
HCP resided in diverse locations both inside and outside of the metropolitan areas the health care systems served, encompassing most of Georgia, Maryland, and northeastern Illinois.HCP worked in more than 100 different physical locations clustered near academic centers (Figure 2).Major infection control practices were generally consistent across sites, although there was some variation between sites and changes during the study period (eTable 3 in the Supplement).

Community Factors and COVID-19 Incidence
Cumulative incidence of COVID-19 in the community, linked to HCP residence by 3-digit zip code 1 week prior to serology testing, ranged from 8.2 cases per 10 000 to 275.6 cases per 10 000.

Regression Model Results
Of the 24 749 participants with determinate serology results, 1201 (4.9%) were excluded because of missing or out-of-state zip codes, leaving 23 548 (95.1%) for the multivariable analysis.Demographic and community factors were associated with higher odds of seropositivity among HCP (Table 2).

Discussion
This study of more than 24 000 HCP found a low SARS-CoV-2 seroprevalence of 4.4% across multiple, geographically diverse health care systems.3][4][5][6]14 Our rate of SARS-CoV-2 seropositivity was similar to that found in some health care centers 3,4 but substantially lower than others. 6In our study, there was no clear association between workplace contact with patients with COVID-19 and antibody positivity, consistent with some studies 1,5 but conflicting with others. 2,3We found that having community contact with COVID-19 increased the risk of being seropositive, similar to another study. 1 While prior studies assessed community exposure primarily through self-reporting, our study found an association between cumulative incidence of COVID-19 in an HCP's residential zip code and seropositivity across a diverse geographic area.We found that the higher the cumulative incidence of COVID-19 until the week prior to the antibody test, the higher the risk of the HCP being antibody positive.This finding aligns well with the observed association that HCP who had contact with a person with COVID-19 in the community were more likely to be antibody positive.Together, these findings suggest that exposures outside of the workplace, rather than exposures to patients with COVID-19, may be major drivers for SARS-CoV-2 infection among HCP in the United States.
A large study in Denmark that included more than 29 000 HCP found that HCP had a higher positivity rate than a contemporaneous comparison group of blood donors and that frontline HCP and HCP with more hospital exposure to COVID-19 patients had a higher risk. 2 Another study across 7 hospitals in Denmark tested more than 25 000 HCP and found a positivity rate of 3.4%, with higher positivity among HCP taking care of patients or working in the emergency department. 3However, similar to our study, a large study of more than 40 000 HCP in New York found no association a Crude OR between the specified factor and severe acute respiratory syndrome coronavirus 2 seropositivity using logistic regression.
b OR for the association between the specified factor and severe acute respiratory syndrome coronavirus 2 seropositivity using mixed-effects logistic regression controlling for all other factors in the table and adjusting for correlation within each health care system (via inclusion of a random intercept).Random intercept had a variance of 0.07 (SD, 0.27).
c Other direct care personnel includes dialysis technician or phlebotomist.
d Other health care professional includes laboratory technician, student, medical technologist, other categories unable to refine.
between work location or direct patient care and seropositivity but did not distinguish workplace and community exposures to individuals with known COVID-19. 6though our study does not eliminate the possibility that workplace exposure to patients with COVID-19 increases the risk of SARS-CoV-2 infection, our findings suggest that for HCP, the risk of SARS-CoV-2 infection from community exposures may exceed the risk from patient exposures, especially considering that these findings were estimated across diverse geographic areas and health care systems.Although we were not able to assess chains of transmission in this study, it is possible that some HCP infections in the workplace were acquired during interactions in non-patient care settings, such as break rooms with other HCP with unknown SARS-CoV-2 infection.This has important implications for strategies for HCP protection.While education to HCP has focused on minimizing patient-to-HCP transmission of SARS-CoV-2, with significant effort appropriately expended to optimize patient triage, testing, and correct use of personal protective equipment, our findings suggest that additional effort is needed to prevent exposures in the community and possible workplace transmission between HCP to preserve the HCP workforce.
Recent data suggest that nurses are among the most common HCP infected with SARS-CoV-2, 15 but this likely reflects workforce demographic characteristics given that nursing is the most common health care role. 16We did not find that certain job roles, including those with prolonged patient contact, such as nursing roles, had increased risk, even though nurses represented the largest portion of seropositive HCP across all job roles in our study.Importantly, these findings suggest that current infection control measures are effective for preventing SARS-CoV-2 transmission when working with patients, and HCP risk of infection may be driven by community and nonpatient care occupational exposures.Prioritizing efforts to practice optimal infection prevention in all health care facilities remains critical to keeping HCP and patients safe and may need to include assessments comparing transmission from patient-to-HCP and between HCP.
Vaccination will be important for minimizing SARS-CoV-2 infection among HCP.Despite similar levels of COVID-19 risk to non-HCP, 14 HCP remain a priority group for vaccination for multiple reasons, including their continual potential exposures in the workplace, the importance of preserving health care capacity, and the risk of transmitting the virus from infected HCP to a large number of at-risk patients. 15Our findings suggest that COVID-19 vaccination and other prevention strategies targeting community transmission will also be critical to prevent SARS-CoV-2 infections in HCP.
Like previous US studies, 4,8,9 we found that Black race was associated with seropositivity, consistent with existing health disparities in the community.Our analyses suggest that the association of Black race with SARS-CoV-2 seropositivity may be owing to existing disparities in community exposure rather than from health care-associated exposures.Surprisingly, we did not find an association between seropositivity and ethnicity, but this assessment was limited by the small number of Hispanic HCP in our study.While most previous serosurveys among HCP have not shown an association between age and SARS-CoV-2, we found that being younger than 30 years, a feature of nearly 20% of our HCP, had slightly increased risk of seropositivity.Younger HCP may be more likely to congregate in groups socially, have children in school or daycare, and have contact with other younger persons who may have fewer symptoms with infection.COVID-19 incidence was also observed to be highest among this US age group during the summer months. 17

Limitations
This study has important limitations.Participants were HCP who volunteered for serology testing and thus represent a convenience sample.Laboratory methods differed across sites, potentially resulting in different overall positivity rates than would have been estimated if methods had been standardized.Questionnaires used at each site were not standardized; however, we were able to successfully map many domains.Risk factors included in our multivariable model were limited to those we were able to map from all sites.For example, we were unable to assess risk associated with participation in aerosol-generating procedures because these data were only available from 3 of the

Figure 1 .
Figure 1.Number of Total Health Care Personnel Participating by Week and Health Care System2500

Figure 2 .
Figure 2. Geographic Distribution of Health Care Personnel in Each 3-Digit Zip Code and Work LocationsGeorgiaA ).Most HCP reported working predominantly in acute care hospitals (21 566 [87.1%]), with much smaller proportions working predominantly in ambulatory settings (1303 [5.3%]) or long-term care or inpatient rehabilitation facilities (630 [2.5%]); the remainder worked in administrative or other locations.

Table 1 .
Baseline Demographic and Key Characteristics of the 24 749 HCP in the Study JAMA Network Open.2021;4(3):e211283.doi:10.1001/jamanetworkopen.2021.1283(Reprinted)March 10, 2021 5/13 Downloaded from jamanetwork.combyguest on 12/17/2023Community cumulative incidence during the periods of the serosurveys was generally higher for HCP at Johns Hopkins and the University of Maryland (eFigure in the Supplement), which conducted surveys later in the pandemic than Emory and Rush.Most HCP (20 072 [81.1%]) reported no known contact with a person confirmed or suspected of having COVID-19 in their community (Table1).

Table 1 .
Baseline Demographic and Key Characteristics of the 24 749 HCP in the Study (continued) a Percentages represent percentage of column.b Percentages represent percentage of row.c Other direct care personnel includes dialysis technician, phlebotomist.d Other health care professional includes laboratory technician, student, medical technologist, and other categories unable to refine.

Table 2 .
Results of Logistic Regression Analyses, Unadjusted and Adjusted, for 4 Health Care Systems With Severe Acute Respiratory Syndrome Coronavirus 2 Serology as the Outcome Variable

Table 2 .
Results of Logistic Regression Analyses, Unadjusted and Adjusted, for 4 Health Care Systems With Severe Acute Respiratory Syndrome Coronavirus 2 Serology as the Outcome Variable (continued) 4 health care system surveys.Because infection control practices were not standardized across all JAMA Network Open | Infectious Diseases Risk Factors Associated With SARS-CoV-2 Seropositivity Among US Health Care Personnel Mapping of 3 Metadata Variables From Site-Specific Surveys From 4 Health Care Systems eFigure 1. Mean and Range for Cumulative Incidence of COVID-19 per 10 000 Among Study Participants Until 1 Week Prior to Serology Testing eTable 2. Results of Logistic Regression Analyses Excluding Data from Emory, With SARS-CoV-2 Serology as the Outcome Variable eTable 3. Results of Logistic Regression Analyses Including a Time Component (Month) in the Model, With SARS-CoV-2 Serology as the Outcome Variable eTable 4. Results of Logistic Regression Analyses Excluding the Random Intercept Component, With SARS-CoV-2 Serology as the Outcome Variable eTable 5. Infection Prevention Practices for Health Care Personnel (HCP) at the 4 Participating Health Care Systems Before (Cells Shaded Yellow) and During or After (Cells Shaded Blue) the Start of Seroprevalence Testing in HCP at That Site