Skip to content

How Risky Is It To Be an Essential Worker?

I got curious a couple of days ago about essential workers. How much more risk of COVID-19 did they face because they had to work through the pandemic instead of staying home or working remotely? Here's an analysis for California workers:

But here's a caveat from a different study of California workers:

California’s high risk workers, compared with those in non-high risk sectors, were far more likely to be immigrant . . . lived in households that were on average larger . . . were more likely to have children . . . and were twice as likely to have multiple families living under the same roof.

I wonder what these risk ratios look like if you control for all this stuff? That is, non-essential workers clock in at a risk ratio of 1.11 while transportation comes in at 1.28. But is that 0.17 difference accounted for by the controls? If so—or even if the controls account for part of the difference—it means that working didn't really add much danger. The only real exception is workers in food and agriculture, whose risk ratio is high enough that it's almost certainly still significant even with controls.

I'd sure like to see this analysis. It seems unlikely on its face that the actual act of working every day played only a small (or zero) role in raising the risk of dying from COVID-19. And yet, these raw risk ratios are mostly not really that big.

I couldn't find a study that controlled for this stuff properly, which also surprised me. Does anyone know if there's one out there?

27 thoughts on “How Risky Is It To Be an Essential Worker?

  1. cmayo

    Why do they need to control for those things?

    If those trades are mostly made up of people who are at higher risk because of their living situations, that means everyone in those trades are at higher risk because... well, they work together.

    Why would you factor out something that's inherent (regardless of the racial equity to this inherent-ness) to a trade when looking at whether a trade is more risky or not?

    Seems like whitewashing to me.

    1. reino2

      The question is whether or not people got COVID from their jobs, including going to and from. If everybody in similar living conditions got COVID at the same rate regardless of their job, then we just answered the question. If we don't know whether or not that's the case, then we haven't answered the question.

  2. D_Ohrk_E1

    Why stop at just controlling for three variables? How about BMI, age, immunocompromised, income, household size, local epi rules, open/closed local economy, gender, race, citizenship status, educational level, past drug use, etc.?

    Hopefully, ppl understand what I'm getting at.

      1. ScentOfViolets

        He's pointing out that the number of confounders runs into the dozens, if not hundreds. At what level of granularity do you stop, and at what cost?

        Of course, silly old me once thought that 'ubiquitous surveillance' meant an unparalleled opportunity to collect medically relevant data. Not any more, of course.

    1. climatemusings

      There are an infinite number of possible control variables. However, only a few would be expected to be highly correlated with the worker type. I'd nominate the top as being:

      age (essential-)
      household size (essential+)
      immigrant status (essential+)
      education (essential+)
      and maybe pre-existing health conditions (essential?)

      with "-" meaning I'd expect that the correlation would lead to less vulnerability for essential workers on average (e.g., they are younger, so less vulnerable) and "+" more vulnerability. And this combines likelihood of infection with likelihood of death given infection.

      (also, as noted below: each category of "essential" workers can be divided into many different subcategories which have differing vulnerabilities)

  3. iamr4man

    The study doesn’t appear to count the risk from violent lunatic Trumpian anti-maskers. There appear to have been a significant number of them.

  4. haddockbranzini

    I would think there'd have been a much wider spread than that. It almost seems like a lot of factors didn't make a difference at all. Maybe it was things like close family gatherings more than anything else. Yes, there were outbreaks tied to workplaces, but the major outbreaks all seemed to happen after large holidays.

  5. reino2

    Controlling for age would be very important. Essential workers tend to be young adults, and it is easier to pick up COVID in old people because they tend to have more serious symptoms.

    1. climatemusings

      I agree that age is very important, but more because age is a significant risk factor for mortality. Since essential workers are younger adults, we would expect that (all other things equal) essential workers should have a _lower_ risk factor than non-essential workers.

  6. Special Newb

    There are poor outcomes aside from dying from coronavirus. Because there's a big element of luck, I don't care about death risk I care about risk of getting infected.

    1. Crissa

      Hard to measure them, though, without socialized medicine to take a large swath in to standardized testing.

  7. Clyde Schechter

    Mr. Drum, your explanation of this data is incomplete and potentially misleading. Checking the study at the link, the graph shows the risk of _dying_, from any cause, during the epidemic period compared to all cause mortality in 2019. In other words, this is *not* a graph of Covid-19 infection risk, it is a graph of total excess mortality. The text in your post implies the opposite.

    This is important because death from all causes is a very "hard" endpoint: it is observed with very close to 100% accuracy. So these figures are going to be much more reliable than they would be if they were Covid-19 infection rates, which are subject to all sorts of distortions by factors that affect access to diagnostic testing.

      1. ScentOfViolets

        Somebody recently said the difference between good Kevin/bad Kevin was the difference between odd and even. Maybe it's something else, something as simple as Evil Dex/no Evil Dex.

    1. Crissa

      The risk of dying from any cause during a pandemic isn't important when 20% more people died while we were supposed to be doing far less risky behavior than normal?

      1. Clyde Schechter

        I don't understand your question. Who said anything about dying from any cause not being important?

        The point of my comment was about Kevin Drum's presentation: he implied it is a graph of risk of getting covid19, but it's in fact a graph of excess deaths. Those are different things. That's all. And, in fact, mortality data is in an important respect better because ascertainment is very consistent and complete.

      2. Rana_pipiens

        "Far less risky behavior than normal" might be overstating it. It might even be wrong. More accidents happen at home than in the work place. People who were able to work from home in particular probably have safer-than-average jobs. Domestic incidents went up. Less commuting meant fewer traffic accidents, but less traffic allowed higher speeds, and fatalities didn't decrease in proportion. Less commuting meant more time for leisure activities, which can be as dangerous as driving.

  8. azumbrunn

    Another problem is the composition of the groups: Who is included in "health care workers"? It stands to reason that nurses in the ICU are at higher risks than those in psychiatric clinics, not no mention personnel like appointment schedulers, insurance billing staff, managerial staff etc.

    ditto for food workers: The risk appears to be concentrated in meat packing operations. Strawberry pickers are likely to be at lesser risk than those. Lumping them into large categories dilutes the effect and obscures any analysis of risk.

    1. rick_jones

      Not just meat packing but any close-in (semi)industrial food setting. Even Amy’s plants were subject to outbreaks.

  9. skeptonomist

    This is the kind of question that won't be answered until we know more about what kinds of situation promote transmission. For example working in a meat-packing plant or a nursing home probably raise the danger considerably. Maybe being a grocery clerk is not so bad. The classifications that Kevin gives are probably overly broad anyway, Certainly groups of people closely brought together indoors with poor ventilation is bad. Such situations could happen in many kinds of occupation.

  10. skeptonomist

    You can't "control" for a factor if you don't know its own effect. Apparently working in certain occupations is highly correlated with certain situations at home (namely crowding), so separating the effects of work and home would not be simple. You would need to find situations where the two things are not correlated. But of course a mathematical analysis should certainly include all relevant factors.

  11. steve222

    I find this pretty believable. It looked all along like most of the spread was among families. So that food worker gets sick and then spreads it to his family. Should probably look at the families of the essential workers to verify. Then again, there have been suggestions that the primary spreaders were those in their 20s and 30s. So maybe those still going out to group activities they should not have, or may be just that they were the ones doing all fo the shopping.

    Steve

Comments are closed.