key: cord-0847163-wdryelqx
authors: Braun, Katarina M; Moreno, Gage K; Buys, Ashley; Somsen, Elizabeth D; Bobholz, Max; Accola, Molly A; Anderson, Laura; Rehrauer, William M; Baker, David A; Safdar, Nasia; Lepak, Alexander J; O’Connor, David H; Friedrich, Thomas C
title: Viral sequencing reveals US healthcare personnel rarely become infected with SARS-CoV-2 through patient contact
date: 2021-04-15
journal: Clin Infect Dis
DOI: 10.1093/cid/ciab281
sha: 8fefff3d2c9c2a89d8d2cb475740d65ae6b5a3d4
doc_id: 847163
cord_uid: wdryelqx

BACKGROUND: Healthcare personnel (HCP) are at increased risk of infection with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). We posit current infection control guidelines generally protect HCP from SARS-CoV-2 infection in a healthcare setting. METHODS: In this retrospective case series, we use viral genomics to investigate the likely source of SARS-CoV-2 infection in HCP at a major academic medical institution in the Upper Midwest of the United States between 25 March - 27 December, 2020. We obtain limited epidemiological data through informal interviews and review of the electronic health record. We combine epidemiological information with healthcare-associated viral sequences and with viral sequences collected in the broader community to infer the most likely source of infection in HCP. RESULTS: We investigated SARS-CoV-2 infection clusters involving 95 HCP and 137 possible patient contact sequences. The majority of HCP infections could not be linked to a patient or co-worker (55/95; 57.9%) and were genetically similar to viruses circulating concurrently in the community. We found 10.5% of infections could be traced to a coworker (10/95). Strikingly, only 4.2% of HCP infections could be traced to a patient source (4/95). CONCLUSIONS: Infections among HCP add further strain to the healthcare system and put patients, HCP, and communities at risk. We found no evidence for healthcare-associated transmission in the majority of HCP infections evaluated here. Though we cannot rule out the possibility of cryptic healthcare-associated transmission, it appears that HCP most commonly becomes infected with SARS-CoV-2 via community exposure. This emphasizes the ongoing importance of mask-wearing, physical distancing, robust testing programs, and rapid distribution of vaccines.

Despite the use of personal protective equipment (PPE) and other strategies to mitigate risk, frontline healthcare workers are at increased risk for infection with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) compared to the general population [1] [2] [3] . Healthcare-associated SARS-CoV-2 infections negatively affect healthcare personnel (HCP) through direct health impacts, lost wages, and secondary consequences for their close contacts [4] . Additional repercussions include staffing shortages, environmental contamination, low morale and other mental health impacts on HCP. Each of these can impact overall quality of care [5, 6] . Here we use rapid viral sequencing and forensic genomics to investigate the likely sources of infection in 95 confirmed cases of coronavirusdisease 2019 in HCP. We further describe how the results of these investigations informed infection control recommendations within a large academic medical system in the midwestern United States.

The US Centers for Disease Control and Prevention (CDC) have released guidelines for infection prevention for HCP interacting directly with patients with SARS-CoV-2 [7] . These guidelines include recommendations for the proper use of PPE, hand hygiene, precautions to be taken during aerosolgenerating procedures, environmental infection control practices and many others. These guidelines, and additional institution-specific infection control measures [8] , were in place at the institution evaluated here. We posit that these guidelines are generally successful in protecting HCP from SARS-CoV-2 infection in a healthcare setting. Here we test this hypothesis using viral sequences collected from infected HCP, as well as concurrent viral sequences collected from the broader community, to investigate possible sources of infection in a series of HCP.

A c c e p t e d M a n u s c r i p t 6 With a few exceptions [9] [10] [11] , viral sequencing is not currently standard practice for investigating healthcare-associated SARS-CoV-2 infections, although we and others have highlighted the potential utility of this approach [12] [13] [14] [15] . It is currently estimated that SARS-CoV-2 acquires ~2-2.5 consensus mutations per month [16, 17] . Viral sequences can therefore be used to infer likely epidemiological relationships. Viruses collected from transmission pairs or from individuals with a shared source of infection are expected to share higher levels of genetic diversity than individuals who become infected at similar times, but from distinct sources. This was especially true during March -December 2020 in the United States, when transmission rates were high and multiple viruses of distinct genetic lineages co-circulated in many areas [18] . By increasing the resolution of inference, rapid viral sequencing can facilitate a targeted approach to examine SARS-CoV-2 nosocomial outbreaks at the level of the individual and the institution, which others have referred to collectively as "precision epidemiology" [19] .

HCP began testing positive for SARS-CoV-2 at a major academic biomedical institution in the American Upper Midwest in early March 2020. We began sequencing viral genomes from residual nasopharyngeal specimens from the individuals involved in these infection clusters. We focused our analyses on HCP infections and infection clusters that were highest risk for nosocomial transmission, as when healthcare-associated transmission could not be ruled out using epidemiological data alone (see methods for details). Each investigation included at least one HCP, all known direct and indirect SARS-CoV-2-positive patient contacts where residual swab was available, and occasionally extended to epidemiologically-linked household contacts.

A c c e p t e d M a n u s c r i p t 7 We consider three potential sources of HCP infection: "patient source" (via HCP-patient interactions), "employee source" (via HCP-HCP interactions), and "no evidence of healthcareassociated transmission". Some HCP infections did not fit neatly into these categories so we have included three additional categories which are defined in full in the Supplemental File 1. These additional categories are "combined patient and employee cluster", "outside community", and "inconclusive". In each category, for us to conclude person A was a likely source of infection for person B, persons A and B must have had known contact with each other, must have been tested within 14 days of each other, and must have been infected with viruses differing by no more than a single mutation [20] .

From 12 March, 2020 to 10 January, 2021 ~1,172 HCP tested positive for SARS-CoV-2 at the institution we evaluate in this study. In total, we investigated 95 HCP (8.1%) and 137 possible patient contacts collected between 25 March and 27 December, 2020 (n=232). Of these, we were able to generate 87 complete HCP sequences and 87 complete patient contact sequences which were used in downstream analyses (n=174). Of the 87 patient sequences, 4 were included in 2 or more outbreak investigations.

We did not find a closely related virus among co-worker and patient contacts in 55 HCP infections.

We identified a specific household or community source of infection in an additional 3 cases (58/95; 61.1%). We find a smaller percentage could be traced to a coworker (10/95; 10.5%) or were part of a patient-employee cluster (12/95; 12.6%). Strikingly, the smallest proportion of HCP infections could be clearly traced to a patient source (4/95; 4.2%). The remaining HCP infections could not be definitively traced to a single source and were therefore inconclusive (11/95; 11.6%) ( Table 1) .

A c c e p t e d M a n u s c r i p t 8 Below, we describe one representative example of three distinct transmission scenarios -no evidence of healthcare-associated transmission, HCP-to-HCP, and patient-to-HCP.

In case #20, we compared the viral sequence of a HCP (HCP 20-1), who tested positive on 5 October, to a patient contact who tested positive eight days prior. A comprehensive caregiver trace of HCP 20-1 revealed a single patient contact with diagnosed COVID-19 (patient 20-A) within the 14 days prior HCP 20-1's symptom onset. HCP 20-1 provided direct care to patient 20-A while wearing appropriate PPE and with no reported lapses in PPE. HCP 20-1 was infected with a virus clustering with the 20G clade whereas patient 20-A was infected with a 20A-clade virus. The sequences of these viruses differed at >20 sites, so we concluded these individuals were unlikely to represent a transmission pair (Figure 1 ).

In case #16, we investigated infections in three HCP who worked in the same department and tested appropriate PPE with no reported lapses in PPE. We generated consensus sequences from HCP 10-1 and nine patient contacts. There was insufficient viral RNA (vRNA) in the remaining six patient contacts to generate high-quality consensus sequences for comparison. The virus isolated from patient 10-G was identical to the virus from HCP 10-1. Given the known epidemiological association between these two individuals, the time separating sample collections (28 July & 5 August), and identical viral sequences, we concluded patient 10-G is a likely source of infection for HCP 10-1 (Figure 3) . However, we cannot rule out the possibility that another patient whose sample could not be sequenced also shared an identical virus.

HCP and patient viruses are broadly distributed throughout a phylogenetic tree showing the diversity of circulating viruses collected from the areas surrounding the academic medical center (Figure 4) . To investigate the possibility that we missed cryptic healthcare-associated transmission, we compared genetic distances between random pairs of healthcare-associated samples against the genetic distances between randomly paired sequences from the community dataset (grey tips in Figure 4 ) within each month in our study period ( Figure 5) . Overall, healthcare-associated pairs do not share substantially greater sequence identity than randomly paired sequences from the community. This is consistent with a relatively limited role for nosocomial spread of SARS-CoV-2. We additionally plot 14 pairs which are very likely to be true transmission pairs based on epidemiological data (e.g. HCP 2-1 and their household contact) and show these pairs are uniformly genetically identical (see dashed magenta lines in Figure 5 ).

The center where we conducted this case series implemented a number of changes to their institutional infection control guidelines based on these sequencing results [8] . The recommendations for extended reuse of medical grade face masks were clarified and now instruct 

HCP across the hospital are involved in caring for people with COVID-19, whether or not they work on an actual COVID-19 ward. With shifting guidelines and PPE shortages that persist today, it is critical to assess the risk that HCP treating people with known SARS-CoV-2 infection will become infected themselves. We used viral genome sequencing to assess the risk that HCP in a large academic medical system treating COVID-19 patients would acquire nosocomial infections. Our results suggest that caring for COVID-19 patients accounted for a minority of HCP infections (n=4). In contrast, HCP at this institution were much more likely to acquire SARS-CoV-2 from infected coworkers (n=10) or outside of the healthcare system (n=58). This result suggests that infection control procedures, consistently followed, offer significant protection to HCP caring for COVID-19 patients in the United States. A similar conclusion was drawn by recent studies evaluating healthcare-associated infections in the Netherlands and in the UK, suggesting this conclusion may hold across healthcare systems [5, 21] . These results are further supported by another recent study which found the most important risk factor for HCP SARS-CoV-2 seropositivity was cumulative COVID-19 incidence in surrounding communities, not workplace factors [22] . M a n u s c r i p t 11 This study has important limitations. We were able to generate high-quality sequence information for a minority of documented COVID-19 cases in HCP (87/1,172; 7.4%) during our study period (25 March -27 December, 2020). Our dataset is therefore incomplete and may not be entirely representative of viruses circulating in this healthcare setting, particularly for asymptomatic cases.

Similarly, we did not sequence viruses from all SARS-CoV-2-positive patients who were treated at the medical center where we conducted this study. Given this limitation, we were often able to exclude patient contacts and co-workers as likely sources of infection in HCP, but we were rarely able to pinpoint the exact source of infection. It is therefore possible we have underestimated the true rate of SARS-CoV-2 transmission in this healthcare setting. However, the finding that randomly paired HCP and patient sequences do not have greater sequence identity than randomly paired sequences from across the surrounding community suggests to us that we have not severely underestimated nosocomial transmission. Our ability to determine the source of infections in these outbreaks was also often limited by incomplete contact tracing data; undocumented exposures between HCP may have occurred inside and outside of the workplace.

This study examined SARS-CoV-2 infections in HCP from a single academic medical center so our conclusions may not be broadly generalizable. However, another recent study evaluated healthcareassociated infections in the Netherlands and similarly found no evidence for widespread nosocomial transmission of SARS-CoV-2, suggesting our conclusions may hold across institutions and healthcare systems [21] . Further, we were not able to differentiate between routes of infection (airborne, droplet, contact) with the limited epidemiological data available to us in this study.

Sampling and contact tracing of nosocomial outbreaks is often coordinated by local hospitals and/or departments of health while expertise in viral sequencing, bioinformatics, and phylogenetics can A c c e p t e d M a n u s c r i p t 12 more often be found in academic laboratories. Successful application of precision epidemiology requires the integration of these areas. This is possible now at academic medical institutions like ours, but presents more of a challenge at smaller, rural, and private patient care centers. Federal support should be provided to help establish and maintain these collaborations in the current pandemic and in anticipation of future outbreaks.

Here we demonstrated how rapid whole-genome sequencing of current SARS-CoV-2 outbreaks in hospitals can be used retrospectively to reconstruct the likely source of HCP infection and prospectively to adjust and improve infection control practices and guidelines. The approach we describe here need not be limited to investigation of pandemic virus outbreaks. Key concepts from genome sequencing and routine pathogen surveillance can be applied to any nosocomial pathogen and inform changes to infection control practices. Overall, while we do find examples of patient-to-HCP and HCP-to-HCP spread, we found no evidence of healthcare-associated transmission in a majority of HCP infections, emphasizing the importance of ongoing measures to reduce community spread through mask-wearing, physical distancing, robust testing programs, and rapid distribution of vaccines. Individuals who reported high-risk community activities, such as attending a wedding, funeral, indoor bar, or plane travel, were also not sequenced. Relevant patient contacts of individuals with no likely exposure source were identified in the Epic electronic medical record using a comprehensive caregiver trace. This function identifies all patient records accessed by a HCP being traced. Diagnostic assays for the samples included in this study were performed in a clinical lab using CDC's diagnostic RT-PCR [23] , the Hologic Panther SARS-CoV-2 assay [24] , or the Aptima SARS-CoV-2 assay [25] .

Detailed descriptions of all infection control measures implemented to prevent transmission of SARS-CoV-2 at the medical institution evaluated here can be found in a recent report by Lepak et al [8] . Briefly, these guidelines include a universal testing policy for all patients, negative air pressure in all locations where SARS-CoV-2 patients are treated, a limit of one visitor or primary support person per patient per day (required to undergo screening prior to entry), establishment of an employee testing site with required employee self-monitoring for symptoms, maintenance of a log of persons entering the room of a confirmed or suspected COVID-19 patient for contact tracing purposes, detailed PPE guidelines, among others.

Detailed methods descriptions can be found in Moreno et al. [26] . Briefly, viral RNA was extracted using the Viral Total Nucleic Acid Purification kit (Promega, Madison, WI, USA) on a Maxwell RSC 48 instrument. Complementary DNA (cDNA) was synthesized using SuperScript IV Reverse Transcriptase [27, 28] . A SARS-CoV-2-specific multiplex PCR was performed using the ARTIC v3 primers [27, 28] .

A c c e p t e d M a n u s c r i p t 14 DNA was made compatible for sequencing using the one-pot native ligation protocol with Oxford Nanopore kit SQK-LSK109 and its Native Barcodes (EXP-NBD104 and EXP-NBD114) [28] . Up to 23 samples, with one no-template control (water), were pooled prior to being run on the appropriate Nanopore flow cell (FLO-MIN106) using the 72hr run script.

Sequencing data was processed using the ARTIC bioinformatics pipeline (https://github.com/articnetwork/artic-ncov2019), with a few modifications. Briefly, we have modified the ARTIC pipeline so that it demultiplexes raw fastq files using qcat as each fastq file is generated by the GridION (https://github.com/nanoporetech/qcat). Once a barcode reaches 100k reads, it maps to the Wuhan-Hu-1 reference (Genbank: MN908947.3) using minimap2. This alignment will then be used to generate consensus sequences and variant calls using medaka (https://github.com/nanoporetech/medaka). The analysis pipeline is available at https://github.com/gagekmoreno/SARS-CoV-2-in-Southern-Wisconsin.

Samples were excluded from downstream analysis if gaps in the consensus sequence totaled ≥20% of the genome. Each sample's consensus sequence was visually inspected in Geneious Prime (https://www.geneious.com) and/or in Nextstrain's Nextclade online tool (https://clades.nextstrain.org/). We used Pangolin's command-line tool to assign sequences to Pangolin lineages (https://github.com/cov-lineages/pangolin).

A c c e p t e d M a n u s c r i p t 15 

Wisconsin-centric time-resolved and divergence phylogenetic trees (seen in Supplementary File 1) were built using the standard Nextstrain tools and scripts [29] . Laboratories responsible for obtaining and genetic sequence data included here, if not our own, are documented in Supplementary File 2. An interactive view of this Nextstrain phylogenetic tree can be found here.

Full length SARS-CoV-2 sequences available on GISAID as of 10 March, 2020 were obtained and filtered on "Wisconsin" and parsed by date of collection into month bins. We used this dataset as a community comparator set. Consensus mutations were called against Wuhan-Hu-1 reference A c c e p t e d M a n u s c r i p t 16 

We gratefully acknowledge Anna Heffron for assisting with sample transport. We also thank all healthcare workers and infection control teams for their ongoing dedication to patient and community health and wellness. We gratefully acknowledge the originating laboratories responsible for obtaining the specimens and the submitting laboratories where genetic sequence data were generated and shared via the GISAID initiative (Supplementary File 2) .

Accession numbers for all healthcare-associated samples can be found in Supplemental File 1. Code to replicate the genetic distance analyses can be found in the GitHub accompanying this manuscript 

Risk of COVID-19 among front-line health-care workers and the general community: a prospective cohort study

Epidemiology of and Risk Factors for Coronavirus Infection in Health Care Workers: A Living Rapid Review

Factors Associated With Mental Health Outcomes Among Health Care Workers Exposed to Coronavirus Disease

Estimating Coronavirus Disease 2019 Infection Risk in Health Care Workers

Rapid implementation of SARS-CoV-2 sequencing to investigate cases of health-care associated COVID-19: a prospective genomic surveillance study

Secondary Traumatization Outcomes and Associated Factors Among the Health Care Workers Exposed to the COVID-19

for Disease Control C, Prevention, Others. Interim infection prevention and control recommendations for healthcare personnel during the coronavirus disease 2019 (COVID-19) pandemic

Implementation of infection control measures to prevent healthcare-associated transmission of severe acute respiratory coronavirus virus 2 (SARS-CoV-2)

Use of Whole-Genome Sequencing of Adenovirus in Immunocompromised Pediatric Patients to Identify Nosocomial Transmission and Mixed-Genotype Infection

Rapid Metagenomic Next-Generation Sequencing during an Investigation of Hospital-Acquired Human Parainfluenza Virus 3 Infections

Use of Whole-Genome Sequencing in the Investigation of a Nosocomial Influenza Virus Outbreak

Application of next generation sequencing in clinical microbiology and infection prevention

Infection control in the new age of genomic epidemiology

Using Virus Sequencing to Determine Source of SARS-CoV-2 Transmission for Healthcare Worker

Serologic Surveillance and Phylogenetic Analysis of SARS-CoV-2 Infection in Hospital Health Care Workers

Measurably evolving pathogens in the genomic era

Estimates of serial interval for COVID-19: A systematic review and meta-analysis

Precision epidemiology for infectious disease control

Temporal dynamics in viral shedding and transmissibility of COVID-19

COVID-19 in health-care workers in three hospitals in the south of the Netherlands: a cross-sectional study

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

Diagnostic Tests for COVID-19

Panther Fusion® SARS-CoV-2 Assay. Available at

Revealing fine-scale spatiotemporal differences in SARS-CoV-2 introduction and spread

Multiplex PCR method for MinION and Illumina sequencing of Zika and other virus genomes directly from clinical samples

nCoV-2019 sequencing protocol v1 (protocols.io.bbmuik6w). protocols.io

Nextstrain: real-time tracking of pathogen evolution

The authors declare no competing interests.A c c e p t e d M a n u s c r i p t 17