Introduction

The first patients with Coronavirus Disease 2019 (COVID-19) in the United Kingdom (UK) were identified in late January 2020 [1]. Thousands of healthcare workers (HCWs) in the UK have since been infected with Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) [2,3]. A report by Public Health England suggested that early in the pandemic, up to 73% of infections in HCWs were due to nosocomial transmission [2]. However, there remains insufficient evidence around key risk factors for infection in HCWs, and particularly what is driving reported ethnic disparities in infection risk. A recent study in the United States of America of over 24,000 HCWs found community exposures to be important in driving SARS-CoV-2 seropositivity but found no occupational predictors of infection [4], whereas a study of the workforce in 1 hospital in the UK found occupational factors to be important predictors of SARS-CoV-2 seropositivity [3]. The specific underlying factors contributing to an increased risk of COVID-19 among HCWs from some ethnic minority groups, compared to White groups, are also poorly understood [5].

We sought to address these knowledge gaps using data from the national United Kingdom Research study into Ethnicity and COVID-19 diagnosis and outcomes in Healthcare workers (UK-REACH) longitudinal cohort study, which is among the largest UK HCW cohort studies and is unique in the richness of its dataset and the ethnic diversity of its participants. Specifically, we sought to determine risk factors for infection in UK HCWs and whether any disproportionate risks of infection in HCWs from ethnic minority groups might be explained by these risk factors.

Methods

This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 Checklist) and the Checklist for Reporting Results of Internet E-Surveys (CHERRIES) (S2 Checklist) [6,7].

Results

Cohort recruitment and formation of the analysis sample

The recruitment of the cohort has been described previously and details, including response rates, are shown in Fig 1 [8,13]. Briefly, 15,119 HCWs started the questionnaire, of whom 1,858 were excluded from the current analysis as they did not provide their ethnicity, and 720 were excluded due to a lack of outcome data. Therefore, 12,541 HCWs formed the analysis sample, 1,769 of whom were not working during lockdown and therefore were not included in analyses of occupational determinants of infection.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. HCW (those in professional healthcare roles or ancillary workers in a healthcare setting or registered with one of the 7 participating UK healthcare professional regulatory bodies—see Methods for a list of participating regulatory bodies).

* There were 15,997 views of the questionnaire, 155 duplicate records were removed leaving 15,842 unique HCW views. † Corresponds to an effective response rate of 57.1% of those who registered/created a profile on the study website (and 84.5% of those who consented, 1.4% of those who were sent an email, and 3.2% of those who opened the email). ‡ Not included in analyses involving occupational variables. A total of 12,402/15,199 HCWs answered the last question of the questionnaire corresponding to a completion rate of 81.6%. HCW, healthcare worker.

https://doi.org/10.1371/journal.pmed.1004015.g001

Description of the analysed cohort

A description of the cohort is shown in Table 1. With reference to the cohort who were working during lockdown (n = 10,772), the majority were female (75.1%) with a median age of 45 (IQR 35 to 54). Approximately 30% were from ethnic minority groups (19.1% Asian, 4.3% Black, 4.1% Mixed, 2.1% Other).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Description of the 2 analysed cohorts.

https://doi.org/10.1371/journal.pmed.1004015.t001

A description of the cohort who were working during lockdown, stratified by ethnicity, is shown in S3 Table. Almost all of the covariates differed by ethnicity. Age was different by ethnic group (p < 0.001), being lower in the Black and Asian cohorts compared to the White cohort (Black 43.5 [IQR 34.5 to 54], Asian 42 [IQR 33 to 51], and White 46 [IQR 36 to 55]). A greater proportion of Black HCWs lived in areas corresponding to lower IMD quintiles than White HCWs. Important differences were also demonstrated when examining religiosity by ethnic group (p < 0.001) with much greater proportions of Black and Asian HCWs describing their religion as being extremely important to their everyday lives compared to the White cohort (41.9% [Black], 19.5% [Asian] versus 5.9% [White]). Ethnic distribution was not equal across regions of the UK (p < 0.001) with a higher proportion of Black and Asian HCWs practicing in London (26.4% [Black], 21.3% [Asian] versus 11.8% [White]), and a lower proportion practicing in Scotland (2.5% [Black], 4.6% [Asian] versus 7.2% [White]) and South West England (4.2% [Black], 5.2% [Asian] versus 10.1% [White]). A total of 62.0% of White HCWs did not have physical contact with COVID-19 patients, this compares to 49.1% in Asian and 49.6% in Black HCWs.

Univariable analysis of risk factors for SARS-CoV-2 infection

Demographic, household, and health risk factors.

Overall, 2,496 (23.2%) of the 10,772 HCWs who worked during lockdown reported evidence of previous infection. Compared to the uninfected participants, the infected participants were younger, with a greater proportion of Black HCWs compared to White HCWs (OR 1.40, 95% CI 1.14 to 1.73, p = 0.001) (Table 2). Comparable patterns of association were seen when including HCWs who reported they were not working during lockdown (S4 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Description of the cohort working during lockdown stratified by SARS-CoV-2 infection status with unadjusted odds ratios for the association of covariates with infection.

https://doi.org/10.1371/journal.pmed.1004015.t002

Occupational risk factors.

The proportion of HCWs with a reported history of COVID-19 was higher among those who reported attending to higher numbers of patients with confirmed/suspected COVID-19 (with physical contact). A total of 16.4% of those that had no physical contact with COVID-19 patients were infected, compared to 40.0% of those that attended to ≥21 COVID-19 patients per week (Table 2).

Multivariable analysis of risk factors for SARS-CoV-2 infection

Demographic, household, and health risk factors.

In the working cohort, older HCWs were less likely to be infected (aOR 0.92, 95% CI 0.88 to 0.97, p = 0.001 for each decade increase in age). HCWs that lived with other key workers, compared to those that did not, had a small increase in odds of infection (1.17, 1.06 to 1.30, p = 0.002). Those who described their religion as extremely important were more likely to report infection than those to whom religion was not important or were not religious (1.28, 1.08 to 1.51, p = 0.004). HCWs who had received at least 1 dose of SARS-CoV-2 vaccine were less likely to have been infected than HCW who were unvaccinated (0.62, 0.54 to 0.72, p < 0.001). Demographic and household risk factors were unchanged if those not working during lockdown were included (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Multivariable analysis of factors associated with SARS-CoV-2 infection.

https://doi.org/10.1371/journal.pmed.1004015.t003

Occupational risk factors.

Compared to doctors, those working in nursing and midwifery roles were more likely to be infected (1.30, 1.11 to 1.53, p = 0.001). The odds of infection were higher for HCWs who attended to a higher number of confirmed COVID-19 patients (with physical contact), with those attending to ≥21 patients per week being 2 and a half times more likely to be infected compared to those who did not attend to any COVID-19 patients. Compared to those who either did not need PPE or reported access to appropriate PPE each time they needed it, those who reported not having access to appropriate PPE at all times were more likely to be infected (1.29, 1.17 to 1.43, p < 0.001). Working in ambulance (2.00, 1.56 to 2.58, p < 0.001) or hospital inpatient (1.55, 1.38 to 1.75, p < 0.001) settings were associated with higher odds of infection, while working in an intensive care unit (ICU) setting was associated with lower odds of infection (0.76, 0.64 to 0.92, p = 0.003), when compared to those not working in these settings. HCWs working in Scotland and South West England were at approximately half the odds of being infected compared to those working in the West Midlands (Table 3).

Association of ethnicity with SARS-CoV-2 infection risk

In a model adjusted for age and sex, there was an increased risk of infection among Black HCWs compared to White HCWs (Fig 2). This association appeared to diminish as more variables were added to the model and, after adjustment for all covariates, differences in odds of infection between Black and White ethnic groups had attenuated.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Shows aORs (indicated by the central icon) and 95% CIs (indicated by the arms with caps at the lower and upper confidence interval) for the associations of the 5 broad ethnic groups (White ethnic group as reference) with SARS-CoV-2 infection and how these changed with sequential adjustment for groups of covariates.

https://doi.org/10.1371/journal.pmed.1004015.g002

Discussion

In this analysis of over 12,000 UK HCWs, we found that nearly a quarter of participants reported having been infected with SARS-CoV-2 within the first year of the pandemic. The richness of the dataset and ethnic diversity of the cohort has allowed us to identify factors that may explain the disproportionate risks of infection between Black and White HCWs. Additionally, we have identified important risk factors for SARS-CoV-2 infection in HCWs, including: working in nursing or midwifery roles, occupational exposure to increasing numbers of patients with COVID-19, lack of access to PPE, cohabiting with another key worker, and working in hospital inpatient or ambulance settings. Those working in particular UK regions (Scotland and South West England) had lower odds of infection than those working in the West Midlands, as did those working in ICU settings.

Our estimate of nearly a quarter of HCWs being infected with SARS-CoV-2 aligns with anti-SARS-CoV-2 seroprevalence estimates in UK healthcare settings, which have been reported to range from 10.8% to 44.0%, varying by UK region in which the study was conducted [3,14–16]. These estimates, including ours, are significantly higher than the estimated seroprevalence in England (prior to the vaccine rollout and after the first wave of the pandemic) which was estimated to be 6% [17], adding weight to the suggestion that UK HCWs are at higher risk of SARS-CoV-2 infection than the general population [17,18].

We demonstrate a strong association between the number of confirmed COVID-19 patients attended to by a HCW and their risk of infection. Previous studies have found conflicting evidence on this point. Caring for COVID-19 patients was found not to increase risk of infection in 2 large studies conducted in the USA [4,19], whereas in the UK, those working in patient-facing roles during the pandemic have been shown to be at higher risk of infection [3,16], and occupational exposure to patients and colleagues with COVID-19 has been shown to increase the risk of HCW infection in UK settings [20,21]. It should be noted that there are different PPE standards recommended by the 2 countries. In the USA, HCWs are advised to wear a “higher grade” of PPE (including an N95 respirator facemask even when not undertaking AGPs) than UK HCWs when attending to confirmed or suspected COVID-19 patients [22] and this may contribute to the differences in infection risk and the significant risk factors for infection for HCWs practising in the 2 countries.

Evidence for the importance of PPE in preventing HCW infection can be found in our analysis, with those who felt they did not have access to appropriate PPE at all times being more likely to have been infected than those who did. Furthermore, those working in ICU settings (where long sleeve gowns and respirator facemasks are recommended at all times) had lower odds of infection than those working elsewhere. These findings are in agreement with existing studies [3,14,18] and, together with the mounting evidence for aerosol transmission of SARS-CoV-2 [23,24], and the suggestion that coughing may generate more aerosols than activities designated as AGPs such as delivery of continuous positive airway pressure [25] support the claim that upgrading PPE standards for HCWs attending to COVID-19 patients (regardless of whether they are performing AGPs) may have a beneficial impact on the risk of HCW infection. It should be noted that the protective effect of working in an ICU setting against SARS-CoV-2 infection is likely to be multifactorial; in addition to improved PPE (when compared to other hospital and community settings), ICU staff may have a reduced number of patients to which they attend compared to staff on general wards (often having a 1:1 ratio of nursing staff to patients due to the intensity of the care requirements), closed circuit ventilators may prevent environmental contamination with SARS-CoV-2 and patients requiring ICU care may be at a later point in the disease course and may therefore be less infectious [26,27].

We found that a higher proportion of those from Black and Asian ethnic groups reported having been infected with SARS-CoV-2 compared to their White colleagues. This is commensurate with other studies conducted both in the USA and the UK [3,4,14,19,28]. Ethnicity is a complex construct; it has been defined as “the social group a person belongs to, and either identifies with or is identified with by others, as a result of a mix of cultural and other factors including language, diet, religion, ancestry, and physical features traditionally associated with race” [29]. Only by a deeper understanding of factors relating to disproportionate SARS-CoV-2 infections in ethnic minority groups compared to White groups can we reduce hospitalisation, ICU admission, and death [30]. One strength of our study comes from the richness of our data, which allows us to determine the contribution that some of these interrelated factors may make to the higher risk of infection faced by HCWs from certain ethnic groups. In the fully adjusted model, there was no difference in the odds of infection between White and ethnic minority HCWs. This does not imply that there is not an increased risk of COVID-19 for ethnic minority HCWs, indeed it has been shown many times (and is apparent in this work in our univariable and age/sex adjusted analyses) that ethnic minority HCWs face an increased risk of SARS-CoV-2 infection compared to their White colleagues [3,14], rather it suggests that some of the covariates included in the fully adjusted model might drive the differences in the odds of infection by ethnicity.

Supporting evidence for this hypothesis can be found in our work as we found there to be an unequal ethnic distribution across other variables associated with increased odds of infection. For example, a far greater proportion of Black participants, compared to their White colleagues, worked in London and a far smaller proportion worked in Scotland, UK regions with among the highest and lowest infection rates respectively. Black HCWs reported far higher religiosity than White HCWs, which was shown to be associated with an increased likelihood of infection in the fully adjusted model. Black HCWs were also more likely to live in areas corresponding to the most deprived quintile and were more likely to have been born abroad than their White colleagues, both factors with a univariable association with higher infection risk. Importantly, our results indicate that sociodemographic and occupational differences between ethnic groups, such as those described above, are likely to be responsible for the increased probability of infection in Black HCWs compared to their White colleagues, as opposed to any innate biological characteristics.

It is reassuring, given the availability of SARS-CoV-2 vaccine to HCWs in the UK, that our study indicates that those who had accepted at least 1 vaccine against SARS-CoV-2 were at lower risk of COVID-19. However, when one considers that other studies have demonstrated clear differences in attitudes toward SARS-CoV-2 vaccination by ethnic group (with lower vaccine uptake and increased vaccine hesitancy demonstrated in some Black and Asian ethnic groups compared to White [13,31]) SARS-CoV-2 vaccination may represent another mediating factor in the relationship between ethnicity and COVID-19, which should be addressed to prevent even greater disparities in infection risk.

To our knowledge, we are the first to find that religiosity (one of the factors interrelated to a person’s ethnicity) is associated with increased odds of SARS-CoV-2 infection. Religion has previously been associated with outcomes from COVID-19 at a population level, with analysis by the ONS showing those from particular religious groups (including Muslims and Hindus) in England and Wales are at higher risk of death from COVID-19 than Christians [32]. The mechanisms underlying our observation are unclear and warrant further investigation.

In our study, current smokers appear to have a lower risk of infection than former or nonsmokers. A recent meta-analysis found that current smokers (compared to never smokers) were at lower risk of testing positive for SARS-CoV-2 but at higher risk of hospitalisation and mortality [33] and a large seroprevalence study in the UK determined that there was a lower seroprevalence of anti-SARS-CoV-2 antibodies among current smokers than nonsmokers [17]. However, there are many factors that complicate the analysis of the effects of smoking on SARS-CoV-2, for example, smokers (because of a heightened awareness of their increased risk of respiratory disease) may be more likely to adhere to pandemic control measures [17], disruption of the mucosal epithelium in smokers may impact upon the sensitivity of PCR assays for acute infection and increase the odds of a negative test in the presence of disease [33]. Furthermore, some observational studies reporting on the effects of smoking on SARS-CoV-2 transmission have been hampered by collider bias, due to smokers being more likely to develop symptoms suggestive of COVID-19 for reasons other than SARS-CoV-2 infection and therefore more likely to test negative [34]. Regardless of the reasons underlying this observation, it is clear that the negative consequences of smoking far outweigh any potential protective effect against SARS-CoV-2 infection.

Risk of infection differed by UK region of workplace with HCWs in South West England, Scotland, and Northern Ireland being at lower risk than those working in the West Midlands. Compared to the West Midlands, these areas have a lower population density and a lower proportion of the population are from ethnic minority groups [35–38], factors associated with a decreased risk of SARS-CoV-2 transmission [39,40]. Additionally, government imposed restrictions aimed at slowing viral transmission differed between the UK nations and this may have influenced the lower infection risk seen among those working in Scotland and Northern Ireland [39].

We found ambulance workers to be at twice the risk of infection compared to those not working in this setting. To our knowledge, we are the first to demonstrate this effect. The reasons underlying this association require further investigation, although may relate to the frontline position of these HCWs and their exposure to the most critically unwell COVID-19 patients, with much of this exposure occurring in comparatively uncontrolled settings outside of hospital. In line with previous work from the UK, we also found nurses/midwives to be at higher risk of infection than those in medical roles [3].

Increasing age was associated with lower odds of infection. This effect may be due to the close correlation of age with occupational seniority. Senior HCWs spend a greater proportion of their time engaged in managerial and administrative responsibilities, and less time engaged in direct patient care, compared to their junior colleagues. This may lead to less occupational exposure to SARS-CoV-2 and therefore lower infection risk [3]. Additionally, older HCWs have been shown to report better access to PPE (which may also be related to reduced patient contact compared to junior staff [41]).

Our study has a number of limitations. There was potential for selection/responder bias, however comparison with the NHS workforce, while not an ideal reference population, indicates our sample is broadly representative, albeit with a lower proportion of ancillary staff (bias in the UK-REACH cohort study has been explored elsewhere [13]). As with any consented cohort study, there is the potential for self-selection bias. Reanalysis with selection weights to account for this bias was considered, and may form the basis of future work using UK-REACH data; however, given the difficulties with determining a reference population (which are due to our deliberately broad inclusion criteria), conducting and validating this analysis was felt to be beyond the scope of the current work. The cross-sectional nature of the study means we cannot infer the direction of causality, since results may be vulnerable to reverse causation and residual confounding. HCWs who thought they had been infected prior to widespread testing and subsequently tested negative for SARS-CoV-2 infection later in the pandemic would be coded as uninfected in our analysis. We may, therefore, have underestimated infection prevalence. Reassuringly, as noted above, the proportion of infected HCWs is in-line with estimates from other UK studies. PCR and serology status are self-reported, although given the implications of positive SARS-CoV-2 tests in a HCW population, we do not expect recall bias to have much effect on our outcome measure. In using multiple imputation to impute missing data, we assumed that data were “missing at random,” while we have no reason to believe that data are “missing not at random” it is not possible to be definitive about whether this assumption holds for every variable in our imputation models, and therefore, it is possible that the use of multiple imputation may have introduced bias. However, the proportion of missing data for each variable of interest are small and results of a complete case analysis are not dissimilar to results obtained from imputed datasets. In our wider analysis of demographic and home factors in the wider cohort, we used workplace region as a proxy measure for area of residence, we acknowledge that this will not be accurate for all participants due to inter-region travel from home to work. Collapsing “access to PPE” into a binary variable and combining “not applicable” with “all/most of the time” could have affected results. Future studies exploring the relationship between PPE and infection risk may wish to separate these categories.

In conclusion, we identified key sociodemographic and occupational factors associated with SARS-CoV-2 infection among UK HCWs in a large national cohort study. These findings are of urgent public health importance, especially in light of the emergence of a highly transmissible variant of SARS-CoV-2 (omicron), against which vaccination may be less effective [42]. The results should inform policies aimed at protecting HCWs in future pandemic waves through individualised risk assessments, proactive vaccination strategies (including the booster vaccines) to those at highest risk, and better communication around drivers of infection risk to safeguard the healthcare workforce. Critically, we demonstrate that Black HCWs in the UK are more likely to contract COVID-19 than their White colleagues. We have identified some factors interrelated to ethnicity that may underlie this association. Further work should focus on examining how these factors might mediate any disproportionate infection risk to inform interventions. This is particularly important given the increased prevalence of SARS-CoV-2 vaccine hesitancy in ethnic minority HCWs [13].

Supporting information

Acknowledgments

We would like to thank all the participants who have taken part in this study when the NHS is under immense pressure. We wish to acknowledge the Professional Expert Panel group (Amir Burney, Association of Pakistani Physicians of Northern Europe; Tiffanie Harrison; London North West University Healthcare NHS Trust; Ahmed Hashim, Sudanese Doctors Association; Sandra Kazembe, University Hospitals Leicester NHS Trust; Susie M. Lagrata (Co-chair), Filipino Nurses Association, UK and University College London Hospitals NHS Foundation Trust; Satheesh Mathew, British Association of Physicians of Indian Origin; Juliette Mutuyimana, Kingston Hospitals NHS Trust; Padmasayee Papineni (Co-chair), London North West University Healthcare NHS Trust; Tatiana Monteiro, University Hospitals Leicester NHS Trust), the UK-REACH Stakeholder Group, the Study Steering Committee, Serco, as well as the following people and organisations for their support in setting up the study from the regulatory bodies: Kerrin Clapton and Andrew Ledgard (General Medical Council), Caroline Kenny (Nursing and Midwifery Council), David Teeman and Lisa Bainbridge (General Dental Council), My Phan and Jenny Clapham (General Pharmaceutical Council), Angharad Jones (General Optical Council), Mark Neale (Pharmaceutical Society of Northern Ireland), and the Health and Care Professions Council.

We would also like to acknowledge the following trusts and sites who recruited participants to the study: Affinity Care, Berkshire Healthcare NHS Trust, Birmingham and Solihull NHS Foundation Trust, Birmingham Community Healthcare NHS Foundation Trust, Black Country Community Healthcare NHS Foundation Trust, Bridgewater Community Healthcare NHS Trust, Central London Community Healthcare NHS Trust, Chesterfield Royal Hospital NHS Foundation Trust, County Durham and Darlington Foundation Trust, Derbyshire Healthcare NHS Foundation Trust, Lancashire Teaching Hospitals NHS Foundation Trust, Lewisham and Greenwich NHS Trust, London Ambulance NHS Trust, Northern Borders, Northumbria Healthcare NHS Foundation Trust, Nottinghamshire Healthcare NHS Foundation Trust, Royal Brompton and Harefield NHS Trust, Royal Free NHS Foundation Trust, Sheffield Teaching Hospitals NHS Foundation Trust, South Central Ambulance Service NHS Trust, South Tees NHS Foundation Trust, St George’s University Hospital NHS Foundation Trust, Sussex Community NHS Foundation Trust, University Hospitals Coventry and Warwickshire NHS Trust, University Hospitals of Leicester NHS Trust, University Hospitals Southampton NHS Foundation Trust, Walsall Healthcare NHS Trust, and Yeovil District Hospital NHS Foundation Trust.

The UK-REACH Study Collaborative group: Manish Pareek (Chief investigator), Laura Gray (University of Leicester), Laura Nellums (University of Nottingham), Anna L Guyatt (University of Leicester), Catherine John (University of Leicester), I Chris McManus (University College London), Katherine Woolf (University College London), Ibrahim Akubakar (University College London), Amit Gupta (Oxford University Hospitals), Keith R Abrams (University of York), Martin D Tobin (University of Leicester), Louise Wain (University of Leicester), Sue Carr (University Hospital Leicester), Edward Dove (University of Edinburgh), Kamlesh Khunti (University of Leicester), David Ford (University of Swansea), and Robert Free (University of Leicester).