Source: Analysis of Virginia Department of Health daily total case counts by age, retrieved via the Wayback Machine.
Et tu, Virginia?
The recent upticks in diagnosed COVID-19 cases in several states all appear to have one thing in common: Younger people. Much younger, on average, than they had seen in recent months. You can read up on that in recent posts here.
So I set about asking and answering this question: Has Virginia seen any such radical shift in the age mix of new COVID-19 cases?
The answer appears to be no. There has been a modest shift toward the younger age categories. But nothing like the wholesale changes that have been reported in other states (e.g., half of new cases were adults under 30).
The graphs above compare the new cases that accumulated in (roughly) the first week of May, versus the most recent week available at this time. This is all based on data posted by the Commonwealth of Virginia Department of Health.
That said, Virginia does put out total case counts, by age group, day by day. As with many of their data products, they do not maintain a history or archive on line. You have to catch that day’s dataset, that day, or it’s gone forever, to the public, unless it was archived. In this case, courtesy of the Wayback Machine, you can recover those day-by-day datasets for the past couple of months.
The top graph is a count of new cases that showed up in those two weeks. It shows that weekly new cases have declined EXCEPT among the very youngest. There, both the 0-9 age group and 10-19 age group have seen slight increases in case counts. All other age groups saw declines in weekly new cases. But you can see that the declines for the 20-29 group were proportionately less than for the older groups.
The bottom graph shows each age group as a percent of the total for that week. Instead of the absolute count, then, this shows the share of that week’s cases. On that graph, you can see that the mix of cases really is shifting toward younger ages. Each age category for persons over age 30 accounted for a declining share of new cases.
I interpret this as showing that cases are skewing somewhat toward younger persons, but that Virginia has not experienced the wholesale shift that is being seen in the states where cases are currently spiking.
Why does this matter?
First, if we do get the kind of uptick that they’re now seeing in (e.g.) Florida and Texas, it is useful to have this background trend nailed down. We’ll quickly be able to see what changed, if anything.
But there are also some fundamental reasons to worry about the age mix of new cases, right now, in Virginia.
Second, I bet a lot of parents would like to know the cause of the continued growth of new cases in the 0-9 and 10-19 age categories. Presumably, you can’t blame that on lack of social distancing in bars (the explanation being given in both Texas and Florida, for the 20-ish population — see prior posts). So, why is that? Illicit playdates? Lack of social distancing on the playground? Day-care-related? Teenagers hanging out? Catching it from the adults in their families? It would be nice if the Commonwealth could shed some light on that, if possible.
Let me emphasize that there have been zero deaths, roughly 100 hospitalizations, and a very low overall infection rate in those age groups, so far, in Virginia. So I’m positively not going for the fear-monger angle on this. I just think that if my kids were still in that age group, I’d sure want to hear that a) somebody was looking into it and ideally that b) they could attribute it to something.
Third, as kids go back to college, the age mix of new cases matters a lot for on-campus spread of COVID-19. Plausibly, 99% of the interactions of the average college student are with other college students. So nailing down the age mix of new cases — i.e., the ages where the disease is currently circulating — seems kind of important.
Fourth, later, as kids return to school, you want to know the odds that new cases are circulating among school children. So, again, you want to know the breakout of new cases, by age.
Finally, I think the age mix of new cases has a significant implication for the size of the “iceberg” of total cases, relative to the diagnosed “tip of the iceberg”. A shift toward younger cases implies that total cases (whether diagnosed or not) are now growing faster than we have previously assumed.
We’ve known all along that we’ve only been looking at the tip of the iceberg when looking at counts of diagnosed cases. Not everyone who gets this is sick enough to see a doctor about it. So, not every who is infected is diagnosed. A typical credible estimate is that for every diagnosed person, there are maybe 4 to 6 undiagnosed cases.
But that’s all based on the age mix of the cases that they’ve seen so far. My guess is that the ratio of total cases (diagnosed or not) to diagnosed cases is probably much larger among younger people than among older people.
My reasoning for this is that, on average, younger people are less severely affected by COVID-19. And so, the likelihood that a younger person would seek medical attention and be diagnosed is lower than for an older person. But I have to say “probably”, because no research actually shows that, directly or indirectly. Presumably, up to now, it hasn’t been an issue, so literally no publicly available research that I could find showed COVID-19 seroprevalence rates by age.
If I’m right about that, then the actual upticks in cases, in the states currently experiencing an uptick, may be larger than they think. As the focus of new cases shifts to a younger population, we’re seeing a smaller fraction of total cases being diagnosed. Inverting that, total cases (diagnosed or not) is growing faster, relative to diagnosed cases, than it was in the past.