Study recruitment, setting, and data collection

As previously described, we recruited HCP from 4 university-affiliated hospitals and associated clinics in Atlanta, Georgia. ^{Reference Howard-Anderson, Adams and Sherman21} Hospitals included a 961-bed, academic, safety-net hospital; a 751-bed, academic, referral hospital; a 529-bed, academic, community hospital; and a 410-bed community hospital. HCP were enrolled over 5 weeks starting in May 2020. At enrollment, HCP had to (1) be >18 years old, (2) have worked ≥1 shift in the previous 2 weeks, (3) have no self-reported COVID-19 symptoms in the week prior, and (4) have no anticipated employment changes. Several COVID-19–specific infection prevention practices (ie, visitor restrictions, universal masking of HCP, symptom and temperature screening at entry, and universal testing of admitted patients) had been implemented at each hospital prior to the start of study enrollment and remained in place throughout the 6-month study period. Universal eye protection during all patient care encounters was recommended ∼2.5 months after the start of enrollment. Designated COVID-19 units were opened and closed as needed based on the volume of COVID-19 patients in each hospital. The Emory University Institutional Review Board approved the study (IRB no. 00000505). This activity was reviewed by the CDC and was conducted consistent with applicable federal law and CDC policy (See eg, 45 CFR part 46, 21 CFR part 56; 42 USC §241(d); 5 USC §552a; 44 USC §3501 et seq).

Participants were followed for 6 months and were provided serum for SARS-CoV-2 serology testing at enrollment, at 3 months, and at 6 months. All visits were completed prior to COVID-19 vaccine rollout. Participants were sent a monthly Research Electronic Data Capture (REDCap) ^{Reference Harris, Taylor, Thielke, Payne, Gonzalez and Conde22} survey by e-mail to report occupational and community activities. Surveys focused on the prior 2 weeks of activity to maximize accuracy of recall with the assumption that these 2 weeks were representative of the entire month. The initial enrollment survey also captured information on demographics, medical history, primary occupational role, and work setting. At 6 months, participants were also asked to summarize key occupational and community activities (Supplementary Table 1 online).

Table 1. Characteristics of Healthcare Personnel and Association With SARS-CoV-2 Seroconversion Status Over 6 Months (May–December 2020)

Note. All variables are no. (%) unless otherwise stated. AGP, aerosol-generating procedure; CDC, Centers for Disease Control and Prevention; CI, confidence interval; HCP, healthcare personnel; ICU, intensive care unit; IQR, interquartile range; PPE, personal protective equipment; OR, odds ratio.

^a Survey options for race included American Indian or Alaska Native, Asian, Black or African American, Native Hawaiian or other Pacific Islander, White, other race, or prefer not to answer. Due to small numbers, American Indian or Alaska Native, Native Hawaiian or other Pacific Islander and Other were combined. We excluded participants who preferred not to answer. Ethnicity was examined separately from race in part because of the small number of HCP that self-reported as Hispanic or Latino (all of whom did not seroconvert) and to maintain consistency with our previously published analysis addressing seroprevalence in this sample of HCP at enrollment. ^{Reference Chan, Ng and Rahman17}

^b HCP were considered immunocompromised if they had an autoimmune or rheumatologic disorder, active malignancy, solid-organ transplant, hematologic stem cell transplant, or other self-reported immunosuppressive condition or medication.

^c Participants who stated they did not have patient care roles were not asked this question on the baseline survey.

^d Excludes HCP where primary location was not able to be determined due to multiple locations being reported.

^e Includes participants who performed no patient care activities or worked zero shifts in the 2 weeks prior to survey completion.

^f Participants without patient care were included in the reference group.

^g The following procedures were specifically included as AGPs: airway suctioning, noninvasive positive pressure ventilation, manual (bag) ventilation, nebulizer treatments, intubation, cardiopulmonary resuscitation, chest physiotherapy, mini-bronchoalveolar lavage, breaking ventilation circuit, sputum induction, bronchoscopy, high-flow oxygen delivery.

^h A high-risk occupational exposure to SARS-CoV-2 was defined based on the CDC guidance as having prolonged close contact with a patient(s) with SARS-CoV-2 infection while (1) the HCP was not wearing a respirator or face mask; (2) the HCP was not wearing eye protection while the patient was not wearing a face mask or intubated; or (3) the HCP was not wearing all recommended PPE (gown, gloves, eye protection, and respirator) while performing an AGP. ³⁹

ⁱ The cumulative incidence of COVID-19 per residential ZIP code was calculated using data from the Georgia Department of Public Health (GDPH) and the US Census Bureau and includes all reported cases of COVID-19 (confirmed and probable) from 1 week prior to enrollment to 1 week prior to 6-month blood draws for each participant.

Serologic testing

A laboratory-developed, indirect enzyme-linked immunoassay (ELISA) detecting IgG antibodies against the receptor-binding domain (RBD) of SARS-CoV-2 was developed and validated similarly to previous descriptions. ^{Reference Howard-Anderson, Adams and Sherman21,Reference Suthar, Zimmerman and Kauffman23} Sera were tested in duplicate at a dilution of 1:100. A monoclonal antibody (CR3022) ^{Reference Yuan, Wu and Zhu24} that binds to SARS-CoV-2 was included in each plate at 200 ng per well as an internal reference to ensure comparability across plates. Longitudinal specimens from the same individual were run on the same plate. The mean optical density of test samples was divided by the mean optical density of CR3022 from the same plate to obtain the normalized ratio (NR). Samples that tested positive on the initial run were considered seropositive only if the result was confirmed on a second independent run. The threshold for detecting SARS-CoV-2 antibodies was determined by receiver operating characteristic (ROC) curve analysis to be NR ≥0.2.

Sample size calculation

This study was designed to detect differential proportions of seroconversion at 80% power with an α of .05, comparing HCP who reported high-risk occupational activities (ie, spending >50% of a shift directly at bedside, working in COVID-19 units or performing AGPs) to those who did not. We assumed an equal distribution between groups and that 30% of HCP with a high-risk occupational activity and 15% without a high-risk occupational activity would seroconvert, corresponding to a sample size of 244. ^{Reference Dean, Sullivan and Soe25}

Variable definitions and statistical analysis

The primary outcome was SARS-CoV-2 seroconversion, defined as having a negative SARS-CoV-2 ELISA result at enrollment followed by a positive result at either 3 months or 6 months. Participants with SARS-CoV-2 antibodies at enrollment were excluded from the primary analysis. Participants with negative serology results at 3 and 6 months, or with a missing 3-month result and negative 6-month result, were categorized as nonseroconverters. Participants who missed both the 3- and 6-month serologic assessments, or who had a negative result at 3 months but missed the 6-month assessment, were excluded from analyses. For the regression analysis, we also excluded participants who were not employed at a study hospital for an entire 3-month assessment period (ie, 0–3 months or 3–6 months).

We examined the following time-invariant variables: age, sex, race, ethnicity, immunocompromised status, primary hospital of employment, and occupation. HCP were considered immunocompromised if they reported an autoimmune or rheumatologic disorder, active malignancy, solid organ transplant, hematologic stem cell transplant, or taking immunosuppressive medications. Participants were classified as administrators, nurses, physicians or advanced practice providers (APPs), or other HCP based on the enrollment survey. Time-variant variables included both occupational and community activities (ie, attending gatherings with >10 people and using public transport) as well as whether participants had close contact (ie, >20 minutes of caring for, speaking with, or touching another person) with individuals with laboratory-confirmed SARS-CoV-2. Because seroconversion information was not measured monthly and because participants could only seroconvert once, longitudinal data analysis methods (ie, repeated measures analysis) could not be performed. Furthermore, while seroconversion was assessed at 2 time points, exact dates of seroconversion were unknown. Therefore, survival analysis was not appropriate for these data. To summarize the time-variant variables across the 6 surveys, we assigned the most commonly reported value for each variable. If no value was most common, then the most common and recently reported value was used (Supplementary Table 1 online).

We used proportions for categorical variables and medians with interquartile ranges (IQR) for continuous variables to characterize the study cohort and to assess changes in occupational activities over time. Univariable logistic regression was used to examine unadjusted associations between all independent variables and SARS-CoV-2 seroconversion. Multivariable logistic regression was used to examine associations between occupational variables determined a priori and SARS-CoV-2 seroconversion, adjusted for potential confounders. The primary occupational variables investigated were average proportion of each shift spent at a patient’s bedside, proportion of shifts worked in COVID-19 units, and performing or being present for ≥1 AGP. We included potential confounders related to demographics, community risk factors, and known SARS-CoV-2 exposures based on directed acyclic graph theory (Supplementary Fig. 1 online). We also included any significant variables identified in the univariable analysis (P ≤ .10) in the final model. In addition to the primary occupational exposure variables, the final model included age, race, immunocompromised status, practicing universal masking at work, having close contact with a SARS-CoV-2–infected coworker or individual in the community, and COVID-19 cumulative incidence per residential ZIP code. The incidence was calculated using ZIP code tabulation areas ²⁶ and COVID-19 surveillance data from the Georgia Department of Public Health and includes all reported cases of COVID-19 from 1 week prior to enrollment to 1 week prior to 6-month assessment for each participant. We considered interactions among the proportion of each shift spent at the bedside, proportion of shifts worked in COVID-19 units, and race. Interactions with race could not be examined due to insufficient sample size. We detected no evidence of an interaction between proportion of shift spent at bedside and proportion of shifts worked in COVID-19 units. All analyses were performed using R version 4.0.2 software (R Foundation for Statistical Computing, Vienna, Austria).

Fig. 1. Results of the enzyme-linked immunoassay (ELISA) assessing SARS-CoV-2 serology status. A participant was considered to have detectable SARS-CoV-2 antibodies if the normalized ratio was ≥0.2 (dotted horizontal line). (A) All 278 participants who did not seroconvert over the 6 months. (B) The 19 participants who had SARS-CoV-2 antibodies at enrollment and were not eligible for the seroconversion outcome. (C) The 26 participants who were seronegative for SARS-CoV-2 at enrollment and had detectable SARS-CoV-2 antibodies at 3 or 6 months.

In a post hoc analysis performed to explore reasons why we did not observe an association between AGPs and seroconversion, we examined the frequency of personal protective equipment (PPE) use among participants who performed or were present for ≥1 of the following AGPs in COVID-19 units: airway suctioning, cardiopulmonary resuscitation (CPR), intubation, and noninvasive positive pressure ventilation (NIPPV).