Post #755: COVID-19, younger people, and explosive growth. A theory, but no data.

Posted on July 13, 2020

Source:  New York Times 7/13/2020  Red and green markings were added for purposes of illustration.

1:  Young people.  The cases that are showing up, in these outbreaks, are far younger, on average, than COVID-19 cases to date.

2:  Explosive growth.  A second outstanding factor in those states was the extremely rapid ramp-up in the case load.  On the graphic above, the green line shows the worst rate of case growth experienced in Virginia.  If you look carefully, you can see that in the three key states, they went from well under that level well over that level in a single week.  And from there, to crisis levels of daily new cases in another week or two.  From “everything is fine” to “public health emergency” in the span of a few weeks.

3:  Relatively few deaths, as shown below.  In part, that’s because deaths lag infections by a couple of weeks, on average.  But these upticks have gone on long enough now that we can also say this is partly “real”, in the sense that the mortality rates do not (now) appear to be climbing as steeply as the new case counts did.

Source:  New York Times 7/13/2020  Red markings were added for purposes of illustration.

The fact that 1) and 3) are related is obvious:  The mortality rate ramps up steeply with age, and COVID-19 mortality case rate among young people is quite low.  I discussed this in Post #730, where I estimated the impact the shift in age mix on average mortality using data from the Virginia.

But I bet that 1) and 2) are also related.  Not due to “bad behavior” by young adults, per se.  (Though that certainly may contribute).  I think low age and explosive growth may go hand-in-hand merely as a matter of arithmetic.

This conclusion is an expansion of the addendum portion of Post #723.  The cases that get counted are just the tip of the iceberg of all infections.  Lately, FWIW, CDC estimated that there are, on average about 10 infections for every reported (diagnosed, tested) case.  (I saw FWIW, because that’s based on a sample of convenience in a handful of sites.)  But I’ll bet that with young adults, the size of the iceberg (all infections) is vastly larger than the tip of the iceberg (cases testing positive for COVID-19). 

And in that situation, you would, as a matter of arithmetic, expect the counted cases to grow much more rapidly.  In other words, for a given uptick in counted (diagnosed) cases, your actual underlying epidemic is much worse if those cases are all young people.  By interpreting the early upticks in these states using  historical norms for a much older COVID-19 population, the states missed out on just how widespread these outbreaks were.

If true, this matters to us here in Virginia, because the recent uptick in cases (see next post) in the Hampton Roads area has been reported to be due to cases concentrated among young adults.  If the theory above is right, then the current case counts underestimate total new infections by a much larger margin than has been historically true.  And so we risk seeing the type of unexpected explosion of case counts as has been noted elsewhere when this has morphed into a young person’s epidemic.

Details follow.


Return briefly to the earliest studies of the COVID-19 outbreak in China.  There, numerical modeling of the outbreak indicated that as much as 86% of the cases there went unreported.  (Discussed in Post #624).

How did they infer that, if the cases were never reported?  The infection spread far too fast to have been spread by the reported cases alone.  In other words, given what the knew about how infectious the disease was, and how fast it spread, they had to make up a vastly larger population of infected-but-unreported individuals to make all the pieces fit together.

Now turn that finding on its head:  The more unreported cases there are, the faster the infection will spread, for a given number of reported cases.  That’s the first key insight.  It’s the size of the entire iceberg that matters, for how rapidly this spreads.  And the smaller the tip of the iceberg (reported cases) is relative to the size of the entire iceberg, the faster this will grow, relative to that tip.

So, for the moment, suppose that younger people have a smaller “tip” (reported, diagnosed cases) relative to “iceberg” (all infections, reported or not).  Compared to what we have seen in the past, for a much older average age mix.  Then, as a matter of logic, for a given size of that “tip”, we’re going to see faster spread of infection, than we would have seen for that same count of diagnoses cases in an older population. For the simple reason that, for young people, that “tip” is actually signalling a much large overall “iceberg” of total infected cases.


The theory above is well-worked-out and pretty much unambiguous.  The only remaining question is whether it’s true that the ratio of total infections to reported infections is materially larger in young adults (say, age 20-29) than in an older working-age population (where the bulk of cases were found historically).

And, more to the point, can I find the data to prove or disprove this?

It’s certainly plausible.  The idea is that cases fall on a spectrum:

  • Asymptomatic — no symptoms ever, no reason to see physician.
  • Moderate symptoms, but not bad enough to require seeing a doctor.
  • Symptoms severe enough to warrant seeking care.
  • Hospitalization
  • Death.

We don’t see the cases in the first two bullets.  We do see the cases in the last three.  Those are the bulk of the persons tested and found to have coronavirus.  Here’s how that looks for Virginia, with question marks for the data we would like to have:

In that key young adult category (age 20-29), only 2% of those who were diagnosed with COVID-19 ended up hospitalized.  Versus 5% for the next-oldest age group. So they were 2.5 times less likely to require hospitalization.  And, while you can’t see it from the data, they were over 3 times less likely to die.

And so, is it reasonable safe to guess that there were more un-diagnosed cases, relative to diagnosed cases, for that group.

The problem is, I can’t find even one data source that will let me fill in that table.  Apparently, nobody has bothered to ask and answer the “size of the iceberg by age” question.

I think this is a fairly key piece of missing information that could be inferred from existing seroprevalence data.  I think it has a material bearing on what we can expect from a pandemic focused on young adults.  But as far as I can tell, this is not something that is public information anywhere.

The closest I can find is a proposed World Health Organization (WHO) study. (“Population-based age-stratified seroepidemiological investigation protocol for COVID-19 infection”)  They started this in Februrary.  Perhaps in a year or two they might get around to publishing some findings.  That would be nice.

But as of today, I can’t find even one study that lets me fill in that table.

If I find it, I’ll post it.  For now, this remains just a theory.  A theory that, if true, may matter greatly as the situation in the Hampton Roads area progresses.