Occupational characteristics associated with SARS-CoV-2 infection in the UK Biobank during August–November 2020: a cohort study

Many severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) outbreaks have been documented in workplace settings [1, 2], and for those doing in-person work, the workplace can be a key point of exposure to SARS-CoV-2 [3]. Most studies of workplace risk for SARS-CoV-2 infection have focused on specific occupational groups with particularly high risk, such as healthcare workers [4‐7]. A smaller proportion of studies have examined a general population sample to investigate occupational characteristics associated with SARS-CoV-2 infection across a diverse array of job types [8, 9]. The UK Biobank, a 500,000 person cohort with linked SARS-CoV-2 testing data, provides an ideal resource for such a study.

One of the first studies of occupational risk for SARS-CoV-2 in the UK Biobank demonstrated that early in the pandemic, essential workers had increased risk of developing severe coronavirus disease 2019 (COVID-19), particularly those in the healthcare and social/education sectors [10]. Other studies have demonstrated that shift work is associated with higher risk of severe COVID-19 [11, 12]. However, these studies had a limited number of occupational characteristics that could be examined, and were conducted whilst the UK was in lockdown (essential work only).

We aimed to leverage a job exposure matrix that was recently linked to job titles collected in the UK Biobank [13] in order to examine associations between several occupational characteristics and SARS-CoV-2 exposure risk. Our study focused on a time period after the UK was released from lockdown when workers in a greater variety of jobs would have been expected to return to in-person work.

The UK Biobank cohort included 502,488 participants at the time of this study, 323,075 of whom reported a job at their baseline assessment visit that could be linked to O*NET information (Fig. 2). Of those, 143,477 (44.4%) were less than 65 years of age and still in follow-up in the UK Biobank by August 2020. Finally, 15,383 health care workers and 12,643 participants enrolled in Scotland and Wales were excluded. The remaining 115,451 people made up the final analytic population (Fig. 2). In this population, 47.8% were male, 91.3% were White, and 38.6% had college or University education (Table 1). People living alone made up 14.3% of the population, while 0.21% had eight or more people in their household.

The distribution of O*NET job characteristics was assessed in the study population. For ‘Exposed to Disease or Infections’, ‘Outdoors, Exposed to Weather’ and ‘Outdoors, Under Cover’ most people had jobs that required less than monthly exposure (O*NET scores < 3) (Table 2). For ‘Physical Proximity’, less than 25% of people had jobs that required working within arm’s length of other people (O*NET score > =4). For all O*NET variables, scores were higher for people with a positive SARS-CoV-2 test, indicating more exposure to diseases/infection, more outdoors work, and closer physical proximity to other people during work (Table 2). Using the Dingel remote work index, 53.9% of the population was identified as having jobs that could be done remotely, while 45.7% of people testing positive for SARS-CoV-2 were identified as having jobs that could be done remotely. Using the Baker remote work index, 23.8% of the population was identified as having jobs that could be done remotely, while 19.2% of people testing positive for SARS-CoV-2 were identified as having jobs that could be done remotely (Table 2). Seventy-four percent of people identified as doing remote work by the Baker index were also identified as doing remote work by the Dingel index (not including those identified as ‘maybe’ doing remote work). However, as the Dingel index identified many more remote work jobs, only 33% of these matched with the Baker index definition of remote work.

In unadjusted analyses, more frequent occupational exposure to diseases/infections, more frequent exposure to outdoors work exposed to weather, more frequent exposure to outdoors work under cover, and working in closer proximity to other people were each associated with a higher risk of a positive SARS-CoV-2 test (Table 3). Having a job that could not be done remotely was also associated with a higher risk of a positive SARS-CoV-2 test, regardless of the classification system used to define remote work. After adjusting for age, sex, race, Townsend Deprivation Index, education, UK Biobank assessment center, household size, and income, all associations between occupational exposures and risk of a positive SARS-CoV-2 test were attenuated, but remained statistically significant (Table 3). For individual O*NET variables, the strongest association was observed for physical proximity with each 1-point increase associated with a 14% increased risk of a positive SARS-CoV-2 test (95%CI: 5, 24%). This was followed by exposure to disease/infections, for which a 1-point increase was associated with 9% increased risk of a positive SARS-CoV-2 test (95%CI: 2, 16%). Having a job that could not be done remotely was associated with 17–20% increased risk of a positive SARS-CoV-2 test, depending upon the classification system (Dingel index aHR = 1.17, 95%CI = 1.05–1.31; Baker index aHR = 1.20, 95%CI = 1.06–1.36; Table 3). Associations were the same with and without adjustment for race (Table 4).

For both outdoors work under cover and outdoors work exposed to weather, associations with a positive SARS-CoV-2 test were attenuated and no longer statistically significant after adjusting for remote work (Adjusted for Dingel Index: ‘Outdoors, Exposed to Weather’ aHR = 1.03 [95%CI = 0.98–1.09], ‘Outdoors, Under Cover’ aHR = 1.04 [95%CI = 0.95–1.13]; Adjusted for Baker Index: ‘Outdoors, Exposed to Weather’ aHR = 1.04 [95%CI = 0.99–1.10], ‘Outdoors, Under Cover’ aHR = 1.06 [95%CI = 0.98–1.15]).

To understand relationships between occupational characteristics and SARS-CoV-2 risk, our study capitalized on a large population-based cohort with a wide variety of occupations during a particularly important time period in the UK when most businesses were open, but vaccines were not yet available. In this population we identified associations between several different occupational characteristics and the risk of a positive SARS-CoV-2 test, indicating that in-person work is an important contributor to SARS-CoV-2 transmission.

Several studies have examined associations between work and risk of severe COVID-19 in the UK Biobank. Early in the pandemic, Mutambudzi et al. found that essential workers, particularly healthcare workers and social/education workers, had higher risk of severe COVID-19 than non-essential workers [10]. Subsequent studies identified similar findings, as well as an association between shift work and higher risk of severe COVID-19 [11, 12]. In contrast to these studies, our study focused on the time period after the UK emerged from lockdown, included occupational characteristics that were not previously examined, and used SARS-CoV-2 positive tests as our primary outcome. However, consistent with these prior studies, our results indicate that in-person work is associated with higher risk of SARS-CoV-2 infection. Importantly, our study demonstrates that this elevated risk is not limited to healthcare workers as this occupational group was excluded in our analysis.

Several of the occupational risk factors we identified were consistent with our initial hypotheses and with studies in other populations. Closer physical proximity and higher exposure to non-SARS-CoV-2 infections and diseases were associated with higher risk of SARS-CoV-2 infection. A prior UK Biobank study by Maidstone et al. identified a positive correlation between SARS-CoV-2 positive tests and a higher work environment score that combined measures for infection exposure and proximity [11]. But the study did not account for potential confounders of this relationship, such as household size or measures of socioeconomic status [11]. A national study in the UK also demonstrated that COVID-19 death rates in 2020 were higher among people working in industries that would typically require more exposure to infections and close proximity to others (ex. taxi drivers and care workers) [27]. Our study also found that having a job that was not amenable to remote work increased risk of a SARS-CoV-2 infection. This is consistent with case-control studies in several other countries, in which telework has been associated with decreased SARS-CoV-2 infection risk [28‐30].

We hypothesized that working outdoors would decrease SARS-CoV-2 exposure, but found that outdoors work was associated with higher rates of SARS-CoV-2 infection. This association was attenuated after adjusting for remote work, indicating that our initial association estimates were partly attributable to the fact that outdoors work cannot typically be done remotely. However, there was still no indication that outdoors work reduced SARS-CoV-2 infection risk. Even jobs that typically require significant amounts of time spent outside, may also include time spent in high-risk indoor settings potentially negating any beneficial effect of time spent outdoors. For example, investigations of COVID-19 outbreaks occurring at construction sites after resumption of non-essential work in the United States revealed that lunches and breaks were frequently taken in indoors areas with poor ventilation [31]. Similarly, police officers have some of the highest O*NET scores for time spent outside, but their job duties may also put them in situations with high-exposure to SARS-CoV-2. An analysis of the UK Office of National Statistics COVID-19 Infection Survey found police and protective services workers to have particularly high risk of SARS-CoV-2 infection compared to other types of workers [32].

Our study had a number of strengths. UK Biobank represents a large sample of the general population in the UK that includes a wide diversity of occupations. We evaluated new SARS-CoV-2 infections during a critical time period in which the economy was open, but vaccines were not yet available to provide protection. Through use of a linked JEM we were able to evaluate a number of job characteristics potentially relevant to SARS-CoV-2 exposure [13]. Finally, we were able to control for a number of relevant sociodemographic variables that could influence SARS-CoV-2 infection risk, such as household size.

Our study also has several limitations that should be acknowledged. First, occupational characteristics were assigned based on jobs that were reported at the baseline assessment visit in the UK Biobank, which may not represent the work participants were doing during the study period. Participants may have changed jobs in the intervening years, or may have been out-of-work due to the pandemic. The bias arising from this misclassification would be expected to bias estimates towards the null hypothesis (i.e. no association) and could lead to underestimation of true associations. Second, we could not account for differences in SARS-CoV-2 mitigation strategies between workplaces. Settings in which workers have more physical proximity or known exposure to people with infections may have more mitigation measures in place. For example, in the healthcare setting, SARS-CoV-2 prevention measures are typically more stringent and healthcare workers were routinely tested for SARS-CoV-2 in the UK during the time period of this study. Our study excluded healthcare workers due to the known differences in COVID-19 testing, but there were likely differences in mitigation measures across the other occupations included in our study. Finally, while our use of a JEM eliminates the recall bias that can be inherent in self-reported job exposures, a JEM only allows us to capture average exposures at the job-level, without capturing worker-level variability. As a result, this introduces non-differential measurement error in our assessment of occupational exposures.

In conclusion, we found that people in occupations that are less conducive to remote work have higher risk of a positive SARS-CoV-2 test. Higher workplace physical proximity to other people and higher general exposure to infections and disease were associated with increased risk of a positive SARS-CoV-2 test. People doing jobs requiring outdoors work did not have lower risk of SARS-CoV-2, highlighting the potential risks of any kind of in-person work. In sum, these findings provide additional evidence that COVID-19 is an occupational disease, even outside of the healthcare setting. While vaccines are now widely available to limit the impact of SARS-CoV-2, waves of infections continue and strategies for mitigating transmission in all in-person work settings will remain important.