Severe Acute Respiratory Syndrome Coronavirus 2 Seropositivity among Healthcare Personnel in Hospitals and Nursing Homes, Rhode Island, USA, July–August 2020

Healthcare personnel are recognized to be at higher risk for infection with severe acute respiratory syndrome coronavirus 2. We conducted a serologic survey in 15 hospitals and 56 nursing homes across Rhode Island, USA, during July 17–August 28, 2020. Overall seropositivity among 9,863 healthcare personnel was 4.6% (95% CI 4.2%–5.0%) but varied 4-fold between hospital personnel (3.1%, 95% CI 2.7%–3.5%) and nursing home personnel (13.1%, 95% CI 11.5%–14.9%). Within nursing homes, prevalence was highest among personnel working in coronavirus disease units (24.1%; 95% CI 20.6%–27.8%). Adjusted analysis showed that in hospitals, nurses and receptionists/medical assistants had a higher likelihood of seropositivity than physicians. In nursing homes, nursing assistants and social workers/case managers had higher likelihoods of seropositivity than occupational/physical/speech therapists. Nursing home personnel in all occupations had elevated seropositivity compared with hospital counterparts. Additional mitigation strategies are needed to protect nursing home personnel from infection, regardless of occupation.

Healthcare personnel are recognized to be at higher risk for infection with severe acute respiratory syndrome coronavirus 2. We conducted a serologic survey in 15 hospitals and 56 nursing homes across Rhode Island, USA, during July 17-August 28, 2020. Overall seropositivity among 9,863 healthcare personnel was 4.6% (95% CI 4.2%-5.0%) but varied 4-fold between hospital personnel (3.1%, 95% CI 2.7%-3.5%) and nursing home personnel (13.1%, 95% CI 11.5%-14.9%). Within nursing homes, prevalence was highest among personnel working in coronavirus disease units (24.1%; 95% CI 20.6%-27.8%). Adjusted analysis showed that in hospitals, nurses and receptionists/ medical assistants had a higher likelihood of seropositivity than physicians. In nursing homes, nursing assistants and social workers/case managers had higher likelihoods of seropositivity than occupational/physical/ speech therapists. Nursing home personnel in all occupations had elevated seropositivity compared with hospital counterparts. Additional mitigation strategies are needed to protect nursing home personnel from infection, regardless of occupation.
but the relative impact in each of these settings has not been simultaneously compared.
The Rhode Island Department of Health (RIDOH) and the US Centers for Disease Control and Prevention (CDC) collaborated on a serologic survey of personnel in hospitals, nursing homes, and first responder agencies (e.g., fire, law enforcement) across Rhode Island. As of July 17, 2020, when the survey was initiated, there were >17,700 persons positive for COVID-19 in Rhode Island, of whom 2,675 were nursing home residents and 1,210 nursing home staff, and just more than 1,000 deaths, most among nursing home residents (14). Because of the disproportionate impact on nursing homes, we made an added effort to include as many nursing home facilities as possible in the survey. This analysis compares SARS-CoV-2 seroprevalence among nursing homes and hospital personnel and assesses characteristics and factors related to seropositivity.

Methods
The serologic survey was conducted throughout Rhode Island during July 17-August 28, 2020. RIDOH performed outreach to all agencies to encourage participation. The protocol was reviewed by CDC human subjects research officials, who determined that the activity was public health surveillance as defined in 45 CFR 46 (15). Participation was voluntary, results were not shared with employers, and CDC did not have access to personally identifying information.
RIDOH provided participating agencies with study information and a link to the secure web-based survey to distribute to employees (Appendix Table  1, https://wwwnc.cdc.gov/EID/article/27/3/20-4508-App1.pdf). Upon completing the screening and questionnaire on a personal device, participants received information about blood collection events at their workplace or nearby facility. Each participant provided 10-15 mL of blood using standard venipuncture techniques. Centrifuged serum samples were transferred to a central laboratory for SARS-CoV-2 antibody testing using the ORTHO Clinical Diagnostics VITROS Immunodiagnostic Products Anti-SARS-CoV-2 IgG Test (https://www.orthoclinicaldiagnostics.com). The emergency use authorization data submitted to the US Food and Drug Administration indicated that this test measures IgG directed at the S1 domain of the spike protein with a sensitivity of 90% and a specificity of 100% (16). Results were reported to participants as negative (signal-to-cutoff ratio <1.0), positive (>1.0), or lack of valid result.
A total of 11,987 participants ≥18 years of age consented to phlebotomy and reported no new symptoms of cough, shortness of breath, fever, change in sense of taste/smell, or positive test for SARS-CoV-2 by reverse transcription PCR (RT-PCR) in the 2 weeks before survey participation. Seven were excluded for lack of valid serologic test result because of lipemia or insufficient sample volume and 1,860 did not work in either a hospital (inpatient units and/or ambulatory clinics) or nursing home. Of the remaining 10,120 participants, 9,863 had occupations in direct patient care and support (Appendix Table 2) and were included in this analysis.
We calculated seropositivity (percent positive for SARS-CoV-2 antibodies) overall and for subgroups. We estimated exact Clopper-Pearson 95% CIs and assessed significant statistical differences by evaluating nonoverlapping 95% CI or χ 2 tests for categorical variables and Cochran-Armitage trend tests for ordinal variables (2-sided with α = 0.05).
We classified participants who reported race/ ethnicity as non-Hispanic Native Hawaiian or other Pacific Islander, non-Hispanic American Indian or Alaska Native, or other race as other race (n = 231, 2.3%) and those who declined to specify race/ethnicity as declined (n = 240, 2.4%). We stratified analyses by primary agency selected by participants: hospital or nursing home. Participants could then choose one or more specific workplaces from a precategorized list or free-text workplaces not listed. Hospital emergency department was inadvertently omitted from the response categories for specific workplace but was included in the analysis based on free-text responses. Some hospital and nursing home participants reported working in additional settings that were not the focus of the analysis (e.g., emergency medical services) or in the other agency type (e.g., 1% of hospital and 2% of nursing home personnel worked in both hospital and nursing home settings). These participants were retained in the analysis, but these other workplaces were reported infrequently and are not shown separately. A precategorized list and free-text option were also provided for occupation. Prespecified categories with low frequencies were combined (Appendix Table 2). Among nursing home occupations, 4 with low sample size were combined (other nursing home: engineer/maintenance staff, pharmacist, receptionist/ medical assistant, and physician, n = 56). Analyzing workplace and occupation simultaneously resulted in small sample sizes. Only occupation/workplace groups with sample size >20 or with absolute 95% CI width >30% were shown to ensure estimate reliability (17). Each workplace was represented as a separate dichotomous variable to allow modeling of non-mutually exclusive categories.
Participants reported the frequency at which they performed aerosol-generating procedures; if they needed complete PPE, as defined by CDC recommendations by occupation and patient contact; if, since March 1, they ever used PPE shortage protocols (extended use, reuse, or both); if they lacked specific PPE components when in contact with a person with suspected/confirmed COVID-19 in the workplace; and if they received training in the previous year on PPE donning/doffing techniques. Participants also reported whether their work involved in-person interaction with the community, patients, or both and if they were exposed (spent >10 minutes within 6 feet) to any COVID-19 positive co-workers, household members, patients, or other persons.
We used generalized estimating equations to model likelihood of seropositivity, accounting for clustering by facility (15 hospitals and 56 nursing homes, using an independence correlation structure). PPE variables had a common category (never use PPE) and were thus collinear. Therefore, only PPE shortage protocol use was included in the model, given evidence that shortages may contribute to transmission (12). Similarly, questions assessing use of individual PPE components had a common category, not applicable. Of these, only use of an N95/powered air-purifying respirator (PAPR) was included in the model, because it had an unadjusted association with seroprevalence. For hospital occupations, physicians were the reference group for comparability to a previous study (18). There were not enough physicians in nursing homes to categorize separately, so occupational/physical/speech therapists were the reference group for nursing homes. No interaction terms were explored. We used SAS 9.4 software (SAS Institute, https://www.sas.com) for all analyses.
Among hospital personnel, 27.2% of those exposed to a household member who tested positive for COVID-19 were seropositive versus 2.4% of those unexposed (Table 3). For nursing home personnel, 54.0% of those exposed to a household member with COVID-19 were seropositive versus 10.9% of those unexposed. For both hospital and nursing home personnel, exposure versus no exposure to a co-worker was associated with higher seropositivity, as was exposure to a patient (with or without PPE use) and exposure to some other person. Seropositivity was higher among personnel with community or patient interaction as part of work responsibilities compared with those without for both hospital (3.2% vs. 0.9%) and nursing home personnel (13.7% vs. 7.3%).
For both hospital and nursing home personnel, we found a significant linear trend of increasing seropositivity with greater procedure frequency of performing aerosol-generating procedures (Table 4). For both groups, seropositivity decreased with decreasing frequency of needing complete PPE. Among hospital personnel, those who reported no shortage of PPE had higher seropositivity than those who reused PPE (p = 0.006). Among nursing home personnel, there were no significant differences in seropositivity between those who reported no PPE shortages and those who reported extended use, reuse, or both. Among all personnel, there were no differences in seroprevalence between those who received PPE donning/doffing training versus those with no training (p>0.05 by χ 2 test). For each equipment type, there were no differences in seropositivity between those who reported having versus not having a specific PPE component, with one exception: hospital personnel who did not have an N95 respirator/PAPR were more likely to be seropositive than those who had this equipment (4.4% vs. 2.6%) ( Figure 3).
In adjusted models ( Figure 4; Appendix Table  3), both hospital personnel ( Figure 4, panel A) and nursing home personnel ( Figure 4, panel B) with exposure to a household member with COVID-19 had the highest odds of being seropositive. Otherwise, seropositivity patterns diverged by facility type. For hospital personnel, older age compared with 18-24 years of age was associated with lower seropositivity and non-Hispanic Black and Hispanic race/ethnicity were associated with higher seropositivity. Among nursing home personnel, there was no significant pattern of seropositivity by age or race/ethnicity. Personnel with work responsibilities including faceto-face interaction with members of the community or patients had a higher likelihood of seropositivity among hospital but not nursing home personnel. Among hospital personnel, nurses and receptionists or medical assistants had a higher likelihood of being seropositive compared with physicians. Among nursing home personnel, nurse assistants and social  workers or case managers had higher likelihood compared with occupational, physical, and speech therapists. Finally, hospital personnel working in surgical units had lower likelihood of being seropositive. There were no associations by frequency of aerosol-generating procedures, use of PPE shortage protocols, or not having or using an N95 respirator/PAPR among either hospital or nursing home personnel.

Discussion
In this study, we compared SARS-CoV-2 seroprevalence among nursing home personnel to hospital per- cases and deaths in nursing homes (12). In this study, the association between facility and community seroprevalence may hold, but with exaggerated SARS-CoV-2 transmission in nursing homes versus hospitals. SARS-CoV-2 seropositivity among nursing home COVID-19 unit personnel was nearly 5 times higher than among hospital-based COVID-19 unit personnel. Nursing home non-COVID-19 unit personnel had seropositivity nearly 3 times higher than hospital general inpatient unit personnel. As of November 17,   19 and prolonged viral shedding, residents who may be asymptomatic or have nonspecific symptoms of infection (e.g., increased confusion), shared caretakers between patients, chronic staffing shortages that may be exacerbated by worker illness, and lack of testing and PPE (10,12,(20)(21)(22). In addition, suboptimal infection control practices have been noted in direct observation studies of nursing home personnel (23). We found patterns among hospital and nursing home personnel that suggest both community-and workplace-acquired infection. In both settings, contact with a COVID-19-positive household member was the strongest risk factor for seropositivity. Adjusted odds ratios for seropositivity by age group and race/ethnicity reflected community patterns (24-26) among hospital personnel but not among nursing home personnel. Other studies have found that seroprevalence was correlated with local cumulative COVID-19 incidence in general (12,13,18). Workplace transmission is suggested by higher likelihood of seropositivity among occupations with frequent and prolonged patient contact or working in common areas: nurses and receptionists/medical assistants in hospital settings and nurse assistants and social workers/case managers in nursing homes. Similar findings were noted in other studies (2,18,27). In hospitals, interaction with patients and community members was associated with higher seropositivity than was having no interaction as part of work responsibilities. Finally, in agreement with results from other hospital studies, our study found lower seropositivity among personnel in a controlled environment: hospital surgical units (5,18). However, in nursing homes, workplace factors appeared to dominate community factors given the elevated risk across occupation and seroprevalence >4 times greater than community levels (2.8%). Intrafacility transmission was found in a study of 2 skilled nursing facilities in which viral strains within each facility were genetically more similar than between the 2 facilities or the community; within 1 facility, there were 2 genetically distinct strains, which suggested community introduction into the facility followed by intrafacility transmission (27). That is, this group of studies suggest that community introduction into nursing homes may result in higher level of intrafacility transmission compared with hospital settings.
In at least 2 ways, the higher seroprevalence among nursing home COVID-19 unit personnel could have been partially driven by cohorting residents. First, even if the probability of transmission in facilities were equal, a higher percentage of infectious patients and residents in COVID-19 units would result in a greater number of transmitted infections. Second, if previously infected staff were assigned to COVID-19 units, seroprevalence among facility staff would be increased through staffing decisions rather than transmission. Without longitudinal or genotyping data, it is not possible to disentangle Unadjusted analyses showed that those with daily requirements for complete PPE were more likely to be seropositive for both groups. However, there were no significant adjusted associations between seropositivity and frequency of requirement for complete PPE or PPE shortage protocol use. These findings suggest that PPE use was likely a marker for increased occupational risk (i.e., frequent close contact with infected patients or residents) and that personnel with the most frequent or intense patient contact may have received priority for PPE supplies or that PPE shortages did not have a major role in transmission in this study. More detailed studies are necessary to disentangle the complex factors surrounding PPE use.
Limitations include the cross-sectional study design. Patient or resident infection status was not ascertained. Infection timing relative to different exposures is unknown. For example, it is unknown whether participants who reported exposure to a COVID-19 positive household member were infected by that contact or introduced the infection into the household. Similarly, among seropositive participants who reported working in >1 workplace, it is not possible to ascertain their contribution, if any, to transmission between facilities. Furthermore, seroprevalence is a cumulative measure; antibody responses are reported to persist for >4 months (28). The extent to which seroprevalence was related to exposures early in the pandemic, when PPE shortages were more acute and infection control measures were still being developed, is unknown. Participation was voluntary among a convenience sample, so representativeness of the population is unknown. However, 56 of 85 nursing homes in Rhode Island were included and seropositivity among nursing home participants was related to resident and staff case counts in facilities, with higher seropositivity with rising quartile of case counts (Appendix Table 4). No information was collected about other possible exposures, such as travel and commuting (e.g., use of public transportation). In addition, there could be uncontrolled confounding, including factors related to other socioeconomic factors, such as less flexibility for household members to telework or otherwise reduce occupational exposures. Strengths included a large sample size that allowed stable estimates among subgroups.
This study highlights the increased risk among nursing home personnel for SARS-CoV-2 infection compared with hospital personnel. Although this study was not designed to pinpoint mechanisms underlying the higher seroprevalence among nursing home personnel, 2 patterns strongly suggest that additional workplace protections may mitigate risk in this setting: the elevated risk among all nursing home occupations compared with hospital counterparts and the weaker signals of community transmission among nursing home settings (i.e., no association between age group and race/ethnicity with seropositivity). Continued attention to adherence with current infection control recommendations (e.g., PPE use, handwashing) and ensuring adequate testing, equipment, training, and staffing are the foundations for bolstering the safety of nursing home personnel (22,23,29).

SARS-CoV-2 Seropositivity in Healthcare Personnel
Data and specimen collection activities and specimen testing were funded by US Health and Human Services (contract no. 75P00120C00036). *The adjusted models were estimated using generalized estimating equations including all variables shown. Bolded adjusted odds ratios and 95% confidence intervals are those for which the 95% confidence interval excludes the value of 1.0. NA, not applicable; NH, non-Hispanic; PAPR, powered air purifying respirator; PPE = personal protective equipment. †For hospital model, physicians were the referent occupation group. For nursing home model, occupational/physical/speech therapists were the referent occupation group. ‡Includes 4 categories with low sample size: engineer/maintenance staff, pharmacist, receptionist/medical assistant, and physician. §Workplace was represented by non-mutually exclusive dummy variables entered simultaneously into the model. Participants in workplaces with sample size <30 or with 0% seropositivity were included in the model but the workplace was not entered into the model. Some participants worked in facilities in the other agency category. That is, 84 hospital personnel also worked in nursing home COVID-19 and non-COVID-19 units, and 239 nursing home personnel also worked in hospital administrative offices, COVID-19 units and general inpatient units. Results for these categories are shown in this table, but not in the main manuscript.