Skip to content

What’s the right metric for closing and reopening schools?

A few days ago I tossed out the idea that we needed to figure out a level of COVID prevalence that, in general, justifies shutting down schools. However, after a brief Twitter conversation with Cheryl Rofer I wondered yet again just what the current COVID prevalence is. That is, how many people are currently infected with COVID?

After looking around for a while I learned that no one even tries to track this and report it. And Mike the Mad Biologist is plenty mad about it. Still, it turns out that a very good estimate is simply the number of new cases over the past two weeks, divided by population. Here it is for every state in the union:

Figures like this are available down to the county level too.

My thought is this: Parents don't want their kids going to unsafe schools. But parents also don't want endless and unpredictable lockdowns based on little more than the best guesses of local officials.

This means that our best bet is probably something rule based. I don't know what the right rule is, but something along these lines:

  • If the local prevalence of COVID rises above x%, then all schoolkids in the county are tested.
  • If the positive test percentage is above y%, then school is canceled. If not, wait a week (?) and do it again if prevalence stays high.
  • When prevalence drops below z%, school is placed back in session.

There may be other data points beside prevalence that are important too. I don't know. But what I'm suggesting is that we should create some kind of simple metric, even if it's imperfect, that guides school closures. If prevalence were reported on a web dashboard, parents could check it daily and prepare themselves if it looked like school closure might be on the horizon.

This would give parents confidence that schools were being closed based on firm numbers, not a finger in the wind. More importantly, it would also give them confidence that closures wouldn't drag on forever. Schools would reopen when the metric says so.¹

I'm writing this suggestion in sand, not concrete. I want conversation. Does a reasonable metric even exist? What do epidemiologists think? Would it be good to adopt one? Would parents like it? What kind of exceptions should we make?

And, of course, note that this is strictly about shutting schools due to widespread COVID infection. It has nothing to do with closing schools for other reasons, such as too many teachers being out sick.

Thoughts?

¹I'm agnostic on the question of what schools do when they're shut down. Maybe they declare a holiday. Maybe they start up remote teaching. This would probably vary depending on local preferences.

35 thoughts on “What’s the right metric for closing and reopening schools?

  1. Total

    “ Still, it turns out that a very good estimate is simply the number of new cases over the past two weeks, divided by population. ”

    Hahahhaheheheh…oh, you’re serious.

    It does? You mean states where there’s little testing going on and states where there’s lots should have their numbers treated the same?

    1. kahner

      yeah, as soon as i read that i though "i hope kevin explains the basis for this claim, because it sounds....unlikely to be accurate".

    2. weirdnoise

      Assuming that positivity rates are known as well as a count of tests done and decent models of contagion and disease progression exist (yes, that's a tall order and there are a number to choose from) it's possible to come up with estimates that at least aren't instantly laughable.

    3. TheMelancholyDonkey

      To a large extent, it doesn't matter much whether the metric is accurate. It matters a lot whether the metric is consistent. If you can develop some confidence that the actual disease prevalence varies consistently with the measured metric, then you can base your closure decisions from the metric.

  2. clawback

    If a strain with very mild symptoms came along people would be complaining about unnecessary closures. Then when a particularly virulent strain became dominant the prevalence standard for closing schools would have to be lowered and people would be whining about that.

    So no, there's no substitute for reliance on good judgment from qualified professionals.

  3. Altoid

    I'm truly sorry, Kevin, it's a worthy idea but can't ever be anything more than a thought experiment. Local districts make the calls in most states but within strictures that in some states forbid them from closures, and in some places the anti-closure pressure groups dominate not only the school board comment sessions at meetings but the school boards themselves in many places. And anything predicated on testing runs into the DeSantises of the world.

    The best we could hope for would be a coherent set of CDC guidelines and recommendations. But-- and it really hurts me to say this, because I really want to think well of them-- they seem unable to decide whether they're in the medical/public health guidance business or in the public policy recommendation business and have caught themselves in a land of semi-incoherence between these stools-- from the sounds of it, too much internal recovery work to be able to present a coherent face to the public, at least for now. Yet more wreckage left by the previous administration.

  4. Steve_OH

    As a single data point, today was the first day back in the classroom at my wife's university, here in Appalachian Ohio. About 7% of her students were out with Covid.

    1. iamr4man

      The elementary school where my daughter teaches in San Mateo County had nearly 10% out with Covid as of last Friday. Haven’t gotten today’s report yet.
      On January 4th we had 45 people hospitalized with Covid. Today that number is 128.

    1. rick_jones

      The Health Board of Hong Kong will be getting their marching orders from Beijing so might as well start at the source.

  5. reino2

    This assumes that there is a reasonable number that can be set. I teach at a school that has a COVID dashboard showing what the current prevalence is. The dashboard still has the prevalence limit that the school set at the beginning of last school year. The limit last year was 50 cases per 100k people in the school district per week. That number was set with significant input from doctors and public health professionals. If we went over that limit, we were supposed to at least consider remote schooling. We are now at 1880 cases per 100k people in the school district in the past week, and we are in person.

  6. Joseph Harbin

    ...how many people are currently infected with COVID?

    According to Worldometer, there are 18,775,000 active cases as of Jan 9. That's based on official reporting, which is severely undercounting prevalence, especially the latest surge. Many cases are asymptomatic. Many symptomatic people are having trouble getting tested. Many people are testing at home and their results are never captured in the data.

    Andy Slavitt:
    A number of experts are writing that the actual number of cases in the US is now 4-5x the reported cases.
    ..A pharmacy chain told me they are selling 1 million at-home tests/day.None counted.
    ...So 600,000 positive reported cases we are seeing is really more like 2.5 to 3 million.
    ...This tidal wave will grow much bigger.
    ...One national security analysis recently said that there could be as many as 10 million positive new cases/day at peak. 4x as high as today.
    ...During January, the number of Americans that will get COVID could reach 100 million by some estimates.

    The "based on official figures" data in the chart above seems way off. California, it says, has 2.4% prevalence. But any realistic view would put that number far higher.

    LAUSD tests all students and teachers weekly (school reopens tomorrow), and for testing last week, 13.5% have been positive. (If that holds, that's about 90,000 to 100,000 who will miss at least one week of on-campus schooling.)

    One thing Omicron has done is break the testing infrastructure. Home kits are sold out. Testing sites are very hard to get scheduled for unless it's a school or company with a contract for employees/students. The "quick" tests are often taking 4 to 6 days to process for results, so the people with positives are often in the blind during their most contagious period. The idea that you can retest 5 days after a positive is kind of a joke. It may be 10 days later before the second test result is in.

    1. rational thought

      I did see that la county school testing positive number too. I take that sometimes as the closest we have to a random sample showing true community prevalence. And that is 13.5%? And mostly for children who tend not to stay positive for as long as adults when infected.

      So what does that imply for actual community prevalence? What percenr would test positive if all were tested? Certainly over 10% but maybe even close to 20%. And might mean that here in LA close to 50% infected per month?

      Looks like omicron is even more contagious than we first thought but even less deadlier too.

      Also know someone severely immune compromised from prior cancer who has tried to be very careful as at super high risk of dying with any infection. She got omicron a few days ago anyway probably at necessary doctor visit. Got sick but did not need hospital and fever broke yesterday and feeling better. And this is someone who had good reason to be afraid of catching a cold .

      1. Joseph Harbin

        One story about Omicron's infectiousness. A family gathering over the holidays, 18 attending, all vaxed (2 or 3 shots), 1 person got sick while there (a prior exposure). In the next 10-14 days, 10 of the 18 got Covid symptons or tested positive.

        I'm one of them. No physical symptoms, but plenty of mental anguish. It sucks.

  7. rational thought

    Realize I posted some things re covid on the prior non covid post .

    Would appreciate if anyone into analysis could read them there and respond. As some data today led me to conclusions which just seem hard to believe and I question if I missed something. And , as I explained there, my estimate of covid risk now is important to me for a real life decision and not just theory. And not sure what to do.

    And bonus for anyone who wonders why I have posted so much last year, I reveal why.

  8. golack

    Did you about the NYC public school student gone viral (posts that is)? They're not being taught this quarter--it's all been covid management....
    https://slate.com/human-interest/2022/01/schools-open-closed-omicron-covid-staffing-shortages.html

    Schools are being kept open for daycare purposes. Some places have been doing fine--for a while--but they are more of an exception.
    The question now is, do we just send the students testing positive home, or do we quarantine their close contacts too?

  9. rick_jones

    Kevin - short of mandatory, random (?) testing of the population, I doubt you can even begin to make a guess as to the prevalence of any given disease. Asymptomatics may or may not be testing. Hell, symptomatic may not be testing, but it seems rather more likely that someone is going to seek-out a test because they aren't feeling well - or they must have a negative result to do something they wish to do.

  10. pjcamp1905

    "If the local prevalence of COVID rises above x%, then all schookids in the county are tested."

    Well, there's your problem. You can't get tests. That's a problem unique to this country apparently.

    1. golack

      Vox had a write up about PR--high vaccination rate, yet omicron overwhelming their systems. Case loads got up to 10x past highs, though hospitalizations just getting to past highs, and ICU usage about half past highs though still going up.

      Now--it really looks like cases are dropping. Unfortunately, the positive test rate is is still rising sharply, even as cases fall...

  11. Greg Apt

    I’m curious how they gauge how many cases there are, it seems that it could be wildly off by many measures. For instance, I just tested positive for Covid between Xmas and New Years. I had a PCR test that came back positive, the company that did my test reported it to the heath officials. I called my work and told them, since I had been in the office they sent out a notification of a positive test by someone in the office. I also messaged my doctor’s office to let them know, that I had only mild symptoms, but that I was monitoring it and would let them know if I needed any care.

    So my question is, how many cases of Covid was that? One for just one person? 3 for each of the different groups that I told (and whom don’t talk to each other I’m guessing?). Do they all know my name so they can cross reference it? If I go with Greg with one and Gregory with another does that get caught?

  12. kkseattle

    In our district, schools are closing when there are no administrators, substitute teachers, or bus drivers available. It has little to do with how many kids test positive.

    And in many of the schools, parents are keeping back 1/3 to 1/2 of the students.

    What are the teachers supposed to do when only half the kids show up?

    1. KenSchulz

      You’re getting to the heart of the problem, which is, what ‘regime’ leads to the greatest learning by the greatest number: in-person, remote, hybrid, platooning? It’s fundamentally a question of effective education, not just of slowing the spread of infection. And local circumstances will differ; things like, in how many homes will an adult be present to oversee a remote-learning day? Have all students been provided a connected device?

  13. Crissa

    Kevin, for every confirmed positive, there are all the people in their home, their carpool, and then their class also knocked into quarantine. And these aren't evenly spread, either, as cases chain and branch among the population. It's not not student in each class, it's one student touched five or ten more, and suddenly they're al out, and there's a teacher missing from the class and...

    Still, bars, museums, etc need to close before schools, lest you just knock the kids into different spaces. Their parents need to be available to take them home. And they can't just be shuttled to other public spaces to spread disease in uncontrolled environments.

    1. Joseph Harbin

      Yep. It's not just the individuals with the disease. We are our webs of relationships, and each case cascades to affect many others.

  14. SC-Dem

    Interesting comments from you all.
    My daughter needed a negative test no more than 2 days before moving back into her dorm. I started working to schedule it six days in advance. We live ten or fifteen minutes away from two towns with populations of about 35,000 and 10,000. We had to go to a town a half hour away to find an open testing slot. This is in South Carolina's Pee Dee region. It is largely rural & small town. It seems to me that what limits testing here is a shortage of testing slots. I've gone looking for home test kits before and never found one in stock at any drugstore.
    I wonder if testing of sewage effluent from schools and communities would be a way to get the kind of data needed for a recommended standard practice of the sort Keven suggests.

  15. jeff-fisher

    This kind of thing could only work if the one true metric was very conservative so it triggered before any of the other things that made keeping school open unwise.

    My local school district just posted a big old list of possible triggers.

    Too many teachers out, too many site administrators out (I think that means people who can be the moment to moment principle for the day), too many of specific staff, too many students, student or staff cases too high and growing super fast, too many individual schools in the district out, too many classrooms out, etc.

    They had specific percentages and soforth, but most items were to consider going remote or closing, more than were a hard close.

    We don't know anything like enough to lay out some simple rule unless it is one that would already have triggered almost everywhere.

    Heck snow days are, I'd say only about 50/50 sane in hindsight failing in both directions. And that's way easier.

    1. jeff-fisher

      Sometimes, most times actually, you have to actually trust the humans put in leadership positions.

      Better hope that choice was made well enough.

  16. D_Ohrk_E1

    Official statistics don't account for all of the at-home tests for most states, nor does it account for the asymptomatic and the mild cold symptomatic who don't test.

    The best metric is a modified (addition of suspected COVID) ILI surveillance. For reference: https://gis.cdc.gov/grasp/fluview/main.html

    Think about it. When SARS-CoV-2 becomes an issue of managing an endemic virus, we should take the opportunity to rethink how we address all serious respiratory diseases. Mitigation should apply whenever there's high (modified) ILI levels.

  17. Vog46

    How about when there's no transportation?
    Of course it should be temporary
    But Greensboro/High Point NC is looking at it differently. Due to driver shortages (not covid related, but affected by it) High school students are being directed to use public buses. Elementary and Middle schoolers still picked up by the yellow school bus.
    This is an interesting idea but exposes high schoolers to the general public that may or may not be held to the same COVID restrictions the students are. They COULD bring it into school with them. If the school has a vaccine mandate the affects would be minimal.
    Or are they just adding to the spread?

  18. quakerinabasement

    "I'm agnostic on the question of what schools do when they're shut down. Maybe they declare a holiday."

    Wouldn't declaring a holiday free students (and their families) to spend more time spreading the virus?

  19. cedichou

    Well, let's look at prevalence in CA, since it's in the middle of the pack. 3%. This means that in a classroom of 30 students, on average one is infected. This means that by sending your kid to school, you are placing in the same room with an infected child.

    Now consider that the CDC prevalence does not include home tests and asymptomatic people who don't know they have the disease and won't test. Say the actual prevalence is twice the official one - not unrealistic.

    Now you have moved from "one in average" (that is, still a decent chance that there is none in one classroom and two in the other...) to "pretty darn sure there is at least one sick kid in every classroom."

    Imagine you are a teacher, and knowing that among your 30 kids, every day, one can give you the disease.

    Or as a parent with younger unvaccinated kids at home, or elderly caregivers to the kids - knowing that there is 100% chance that they will get covid over the course of the year because you are playing Russian roulette every day. Sure, 1 kid in 30 may not contaminate your child today, but do it every day of the year...

  20. Jimm

    Track hospitalizations, not cases. With schools, keep tracking attendance and sick days, and have a "zero tolerance" for cold/flu symptoms in schools (kids can catch up and still work on their lesson plans for those few days, and there really should be the capability to be in-class or remote (mixed), at least for the rest of this school year (and be prepared to do it next).

Comments are closed.