From Stat, "Health care workers in the UK are seven times likelier to fall ill with severe Covid-19 than the rest of the population, according to a large new study." The statement is factual; it's just without a context. It is far scarier than need be.
Here is the remaining portion of that quote.
"Researchers compared data from more than 120,000 middle-aged essential and nonessential employees, and found that despite only making up about 10% of the sample, health care workers were seven times likelier to have severe Covid-19."
There are several lessons and insights we might take from the article, but I am not sure this is the critical one; it may be more "if it bleeds, then it leads."
Here are the study's details. It is based on the UK Biobank, a longstanding repository of health information that typically researches genetic associations or genome-wide association studies (GWAS). I guess with some spare time, their interests have turned towards COVID-19. In any case, they used their repository of health information to look at roughly 120,000 individuals – age 49 to 64 and employed. Their baseline information allowed them to categorize each individual's occupation as essential or non-essential. Limitation 1 – the occupational data was collected several years ago, so that the categorization may be a little iffy, but we will spot them accuracy.
30% of participants were classified as essential. Limitation 2 - The UK Biobank dataset participants are a biased reflection of COVID-19 demographics, under-representing both ethnic minorities and low-income groups. This, at a minimum, makes the conclusions land generalization of their findings to groups that are at greater risk an exercise in problematic extrapolation.
The outcome of interest was "severe COVID-19," but what does that mean? For the study's purpose, it was a positive COVID-19 test in a hospital setting. There is a lot to unpack here. Despite a careful search, the type of test is never identified, so the reliability to define an active case of COVID-19 is unclear, especially with indirect tests searching for antibodies. "Hospital setting" included the Emergency Department, where people can and were sent home. Therefore, the numerator described as severely ill includes patients who were not so ill as to be admitted to the hospital along with those sick enough to require hospitalization – apples and oranges.
The other difficulty is the denominator.
"Participants testing negative or positive outside a hospital setting were included in the denominator. We were not able to identify asymptomatic or symptomatic cases who did not present to the health service, and therefore these were also included in the denominator."
I suppose when you mix apples and oranges in the numerator, it is not a problem to continue the mixing in the denominator – perhaps those numbers can sort themselves out. Limitation 3 – the numerical results have uncertainty that expressing them as numbers does not convey.
Of the 120,000 participants 3,111 had COVID-19 testing, with 271 or 0.2% being classified as severe . Of all the non-essential workers, 0.1% were admitted for COVID-19; for healthcare workers, a subset of the essential, that percentage was 0.9%. From there, the researchers provide several other ways of cutting up the data by categories or relative risk.
It is that last measure that is being widely reported. Relative to the risk of non-essential workers, health care workers were seven times more likely to develop "severe COVID-19."  Of course, the other way to have shared their results would be to describe the absolute difference in risk – in this case, 0.8%. Which number bothers you more, a seven-fold increase or 0.8% - I'm betting, and the authors are betting that the relative risk gets your attention.
Another consideration, generally, in looking at numbers is to identify a broader context. During a similar period, the hospitalization rate for this age group in the US was 20%, as determined by the CDC and reported by Vox. That makes the overall incidence of severe COVID in this study 100-fold less than our CDC reports. It's not that one number is correct and the other not; it may merely be we are again talking about very different groups.
A careful look at the study would suggest that its limitations make the reliability of their findings uncertain and not transferable to any other grouping. I don't see evil intent; I see a careful framing of the data. To my mind, the study adds little to our understanding. The problem is the phrase, "seven times likelier to fall ill with severe Covid-19 than the rest of the population," gets your attention while at the same time cluttering your mind with a meaningless meme to pass along.
 Even this number is a bit off, as nine patients had COVID-19 listed as a contributory factor in their death and were tossed into the pile of those admitted for COVID-19
 It isn't nine times more likely due to various statistical adjustments.
Source: Occupation and risk of severe COVID-19: prospective cohort study of 120,075 UK Biobank participants "Occupational and Environmental Medicine DOI:10.1136/oemed-2020-106731"