Post #786: Coronavius odds, the college version

Source:  CNN

As the night follows the day, if there’s a gory school bus wreck anywhere in the USA, you can bet you’ll read about it.  Not because it has anything whatsoever to do with you.  But because it triggers all those hard-wired parental synapses.  News providers know that, and exploit that to the hilt.  You can’t help but click and read the story.

Today’s Washington Post had an article about coronavirus problems at colleges and universities that have resumed in-person classes.  As is typical for articles of this type, the purpose was to find the worst outliers and use them as click-bait to get people angry and upset.  And, judging from the comments under that story, I’d say they did an excellent job of that.

But if you’re a parent with a kid at college or soon going off to college, that sort of analysis is the last thing you need.  What you actually need is some estimate of the odds.  Just how dangerous is it, really, to send your kid back to college?  So in this article, I work through that bit.  Mostly for my own peace of mind.  But possibly for yours as well.

If you want to shortcut all of this, if you read that Post article, be sure to read the one currently shown below it.  Detailing how this was not a problem for British schools in their summer term.

To cut to the chase:   What I found, in looking at a few Virginia universities, is that they fall into three groups.

  1. Some, like William and Mary and Virginia Tech, appear to be testing everybody.  No surprise, when you test everybody, you find a fairly large number of cases / 100,000 persons / exposure day.
  2. Others appear to be testing almost nobody, or testing symptomatic individuals only, and they find far fewer case / 100,000 / day.   So they get “better numbers”, but I’d say that’s a vastly inferior policy for containing any outbreaks.
  3. Yet others report nothing about current campus testing rates and number of COVID-19 cases.  They flunk my test.

The upshot is that you can’t look at the raw rates and form any particular judgment for many Virginia universities.  Some are testing every student, more or less.  Some are testing only those who appear symptomatic.

In any case, none of the apparent infection rates appeared very high.  To me.  As a parent of a student who will soon be returning to her college campus.  For that that’s worth.

The bottom lines here are the following.  The Post article is more-or-less just straight-out fear mongering.  And if you check out the data for your child’s university or college, the first thing you have to ask is what fraction of students are being tested.  You have to know that before you can judge any information about the number of cases they’ve found so far.

If you want it totally boiled down:  Best I can tell, it’s not appreciably riskier, right now, for kids to be off at the typical Virginia university, compared to living in the community.  There’s some slack in that judgment, because I’m looking at tiny differences between very small numbers.  And that may change.  And we should never underestimate the potential for grossly stupid behavior to occur.  But that’s how I see it, as of now, given the available information.

Edit:  It looks like Radford may be Virginia’s exception, and our home-grown outlier.  I’m not sure how the Post missed the opportunity to report that.


There’s no denying the fundamental risks

In terms of the risk of transmitting coronavirus, colleges pretty much tick all the boxes.  Living in group quarters is a risk, for sure.  That’s why we see outbreaks in prisons and jails.  Crowded indoor parties are surely a risk, and are the initial events for many superspreader events.  Going to bars definitely appears to be a risk, which is why state reopening plans have tended to open the bars last, and shut them at the first sign of trouble.   And, to some degree, indoor restaurant meals are a risk.

Add that up, and having a whole bunch of college kids in one place absolutely is a risk.  Particularly if they are drinkers.  To such an extent that I sincerely hope that Virginia universities and colleges are arranging appropriately distanced outdoor events for the students, to divert any pressure for students to organize their own indoor parties.

And it’s undeniable that some students will engage in risky behavior of the type mentioned above.  Which, selfishness aside, is completely rational, given how little risk of serious harm they themselves face from coronavirus.


Baseline risk, living in the community.

Just how little risk of serious illness do college-age people face, from coronavirus?  Let me start by showing their risks, right now, before heading off to college.  Let me take the data for all Virginia residents, for the past month, and calculate the odds. The numbers below are all taken from Virginia Department of Health Dashboard (except population, which is from the US Census).

Depending on what age group you think best matches your child, right now, as a community resident, the odds of serious illness (hospitalization or worse) amount to 2 in 100,000 to 5 in 100,000, per month.  And that’s for all residents, of all socioeconomic classes, work statuses, and living situations.  But because infections hit the poor, workers, and institutionalized persons much more than average, the odds for (e.g.) a well-to-do suburban kid ending up hospitalized with this are easily less than half the odds shown.

To put these rates in perspective, what is the likelihood of hospitalization for all causes, for this age group, in Virginia?  The only data I can find shows non-obstetric admissions for the 1-17 age group (here, .pdf).  Based on that, the odds that any Virginia resident age 1-17 would be hospitalized, per month, is about 0.09%, or 90 in 100,000 residents per month.   Or something like 18 to 45 times higher than the likelihood of being hospitalized for coronavirus.

How much did you worry about your son or daughter being hospitalized, last month, for some random illness?  Well, then, you ought to have worried about them being hospitalized for coronavirus maybe 1/20th as much as that.  And if you never even gave that a thought?  Well, then, there’s your baseline as to how much you should have worried last month about your kid, coronavirus, and serious illness.

Finally, for use below, let me restate one of the rates to use as a benchmark.  The 10-19 age group works out to 9 infections / 100,000 / day.


Now about that Washington Post article.

First, to put the reporting in perspective, what is the universe of US colleges and universities?  That’s a somewhat slippery number, but I’ll take the US News and World report estimate of around 4,300 degree-granting post-secondary institutions in the US.

And so, out of about 4,300 targets, the Post managed to find about a half dozen with apparent problems.  The word for this is outliers.  Out of a universe of 4.300 colleges, the Post reported on a half-dozen apparent outliers.

Second, note that almost all the colleges referenced in the Post article are very large.  I looked up the enrollments and most were in the 30,000 to 40,000 range.  That’s a dead giveaway that the Post reporters were looking for big numbers (clickbait), and didn’t actually look for places where a large percentage of the student population had gotten infected.

Let me just take one example and reduce it to a more meaningful number of new cases / 100,000 population / day.  This is what you commonly see reported for individual states.  For example, Virginia’s 1000 new cases per  day works out to about 11 new cases / 100,000 population / day.

The posts reports that University of Iowa has uncovered 130 cases of coronavirus in the last eight days.  But the enrollment is about 33,000 persons.  To which you would about 10% for teachers and other staff.  When I do the math, I come up with 45 cases / 100,000 / day. Of which, consistent with the odds, there appear to have been no deaths and no hospitalizations.

But what’s the current infection rate for Iowa as a whole?  Based on the most recent NY Times coronavirus map, its 20 cases / 100,000 /day.  So a slightly more objective  headline might have read something like:  Reported coronavirus infection rate at Iowa state is over twice the state average.

But even then, I’m going to spell out everything that’s wrong with that, below.

Third, consider the timing.  In the community it takes maybe 12 days between infection and reporting.  That’s an average of five days between infection and onset of symptoms, and another seven days for the person to get to the doctor, get tested, and have the test results reported.  Most of the universities in the Post report haven’t been open that long.  My conclusion is that much of what they are seeing is due to intensive testing catching pre-existing cases (despite testing prior to coming to campus).  And not necessarily students who were newly infected at the university.

Fourth, consider the impact of greatly enhanced testing.  These student populations are being tested far more intensively that the community population.  Given the very large number of cases that apparently go unreported in the community (something like 10 cases unreported for every reported case, per the US CDC), more intensive testing may find far more cases — that would not have been found, even if they had occurred, if the individual had remained a resident in the community, and not gone off to college.


What about Virginia colleges?

Note the absence of any Virginia colleges in that Washington Post article.  You can be absolutely sure that if the Post could have pulled your chain by bringing up some local university, they would have.  So that’s a pretty good indication that there are not (yet) any huge problems at Virginia universities.

In fact, most colleges and universities in Virginia already have a COVID-19 dashboard where they report the rate of infections.  So if you want to have a look at a cross-section of local universities, it’s not a lot of effort to do so.

What I found, in doing that, is the Virginia universities immediately fall into three groups.

  1. Some, like William and Mary and Virginia Tech, appear to be testing everybody, and they sometimes find far more cases / 100,000 / day.
  2. Others appear to be testing almost nobody, or testing symptomatic individuals only, and they find far fewer case / 100,000 / day.
  3. Yet others report nothing about current campus testing rates and number of COVID-19 cases.

Kudos to William and Mary and Virginia Tech, even if their strategy does make their apparent infection rate look far higher than the (non-testing) other Virginia universities.  I don’t know about you, but if I had a choice, I’d send my kid to a university that tested everybody, even if it does make the numbers look worse.  For a group-quarters living situation like a university, that makes far more sense than testing only symptomatic cases or only a tiny fraction of students.

A few bits of detail follow.  Not an exhaustive list, just enough to give you the gist of what’s going on:

George Mason.  For whatever reason, they don’t have this live, only as a .pdf accessible on this page.  They don’t appear to be testing very many people (like a hundred per week), so take this with a large grain of salt.  So far they’ve found 5 known cases among students, with an enrollment of about 30,000.  That’s for three weeks of exposure.  It’s tough to say exactly what that means, as they don’t randomly test many people.  And I think they still have a lot of commuter (non-live-in) students.  But at face value, that’s a rate of 0.8 cases / 100,000 / day, vastly below the Virginia benchmark (above) of 9 / 100,000 / day.

Cynically (and incorrectly) I could very easily conclude that your kid is safer at George Mason than in the community.  (That’s not true, because this is a highly self-selected population.  Likely their risk in the community was far below average.)  But I think that’s mostly because they’ve tested very few students.

William and Mary has a live COVID-19 dashboard at this link.  Judging from the numbers, they appear to be testing everyone, or very close to it.  In the last seven days, they tested 1600 people and found four positives.  (Their total enrollment is about 8600, but only first-year students are back on campus so far).  Just-flat-ignoring the problem that small numbers are unreliable, that works out to 35 /100,000/ day.  (But with just four cases, the true rate could be almost anything.)

I think the right way to say it is that an exhaustive testing regimen at William and Mary reveals an apparent high new infection rate.  A new infection rate that is four times the community average.  But my guess is, that’s because they seem to be testing literally everyone. 

For the University of Virginia, I couldn’t find a dashboard, possibly because one of their research arms maintains a world dashboard of sorts.  So I don’t know if there is any public information in U.Va.

The James Madison University dashboard is at this link.  They are only testing symptomatic individuals.  They report 4 known cases, out of 616 tests, since July 1.  Tough to make anything out of that.  Per all students (with more than 20,000 enrolled) that’s a miniscule rate.

Virginia Tech has its dashboard at this linkThey appear to be testing a very large fraction of students, with more than 9000 tests (and a total enrollment of about 31,000.  They’ve had 21 positives out of 9,000 tests since August 9, or something like 14 / 100,000 / day.  As with William and Mary, that high rate is consistent with a policy of testing everyone.


Radford

Source:  Wayback Machine view of Radford University COVID-19 web page, Radford University COVID-19 web page.

Radford has its dashboard at this link.  It’s a little hard to interpret their numbers.  With an enrollment of about 9400, they’ve tested 1700 cases, found 5.75% positive, for about 100 cases, total.  But that’s a combination of “move in testing”, and ongoing testing of  symptomatic cases and their contacts, only.   Most other universities required testing prior to moving to campus, they kept any positive cases off campus, and they did not include those pre-move-in tests as part of their counts.

Because they include move-in testing, we’d expect to see their rate be somewhat higher than other universities.  That said, post-move-in, they are only testing symptomatic individuals and their contacts, per their dashboard.  The upshot is that in the post-move-in period, you’d expect a fairly high positive test rate, but not a lot of people counted, because they’re not testing everyone.

Bottom line is that it’s tough to interpret the information that Radford is providing.  That said, a known 100 cases out of 9400 enrollment, if that can be viewed as occurring over 23 days, puts them at a rate of 44 / 100,000 / day.  That’s well above the background community rate.  But because you can’t tell how many they collect upon move-in, versus the number of cases that newly emerged among students post-move-in, it’s hard to say whether they have a major problem or not.

If my kid were going there, I’d check that dashboard and see if the numbers climb over time, or not.  To a limited degree, we can use the Wayback Machine to look at the numbers as they were posted on 8/20/2020, the first time the Wayback Machine captured that information.  That’s shown in the first graph at the start of this section.

And, OK, they plausibly have a problem there.  But I cannot say for sure.  As of August 14, they had 1470 tests and a 0.75% positive rate.  That’s 11 known cases.  That date is the end of their move-in period, but it’s not clear what the lag in their test reporting is.  Then, as of August 23 (9 days later), they had 1705 cases with 5.75 positive, for 98 cases.  Subtract, and in that 9-day span, they identified 87 new cases, or roughly 10 cases/day.  (Above are images from their website).

Again, there’s some uncertainty as to when the move-in cases we counted.  By eye, their 0.75% positive rate in the first graphic just above is a good match for the William and Mary pre-move-in positive rate (12 cases out of about 2000, or about 0.6% of W&M first-year students, tested positive in the week before move-in).

So it’s not unfair to guess that the change between the slides represents new cases.  If so, 10 / day / 9400 works out to 100 / 100,000/ day (rounded), which is fairly self-evidently an excessive rate of new infections, compared to other Virginia universities, if that true is new infections while on-campus.  Which we cannot tell absolutely for sure, from the posted data.  And which remains the major uncertainty there.

This should also be tempered with some notion of the lag between infection and reporting.  If everybody moved to campus on the 14th, we really should not be seeing the new infections in the data yet.  If everybody moved in a week earlier, on average, then yes, we probably should.

Let me put it this way.  That number seems almost too high to be plausible.  But if it’s true, then Radford is far out of line with other Virginia universities, and the Radford administration needs to recognize that and address it in some fashion.

Late edit:  Radford University  is aware that there’s a problem and has started placing additional restrictions on gatherings, per this news article.  About half the population of the City of Radford is Radford U. students, so whatever the university does will have a profound impact on both case counts, and the actual spread of infection in the City of Radford.  It’s now a question of whether actions by Radford U administrators are adequate to bring their numbers in line with other Virginia universities.

If you’re the parent of a Radford student, it’s well worth your while to read that news article cited above.  The Commonweath’s public health official for that district (a Dr. Noelle Bissell) was quoted as saying: 

“In general I’ll tell you the students who are positive, are minimum to non-symptomatic,” said Dr. Bissell.

I’m not quite sure what to make of that, but she has no direct link to the University.  In any case, her characterization is consistent with the Commonwealth’s hospitalization data for the City of Radford, which indicates that nobody has been hospitalized for COVID19 that city since August 5. 


Note:  A key technical point to keep in mind is that tests have a false-positive rate.  That is, some tests are going to show positive, even if the person is not infected.  I’m not sure about the extent to which these universities re-test their positives to confirm them, but throughout this, I am assuming that they do.  If not, the actual reported infection rates at many of these universities could be more-or-less zero.  Many are within what I would plausibly consider the false positive rate for the currently available genetic (“PCR”) tests for coronavirus.