The true shape of Covid-19 pandemic may surprise you
Antibody surveys have made it possible to estimate how Covid count was underestimated earlier
How many people in the US have contracted the coronavirus that causes Covid-19? According to the semi-official confirmed-case numbers gathered from state governments by the Covid Tracking Project, it's a bit over 13 million.
Everybody knows that number is an undercount. Early in the pandemic, limited testing capacity meant that only a small percentage of Covid cases were recorded. Antibody surveys have since made it possible to estimate just how many were missed: In New York City there were 174,504 cases reported in March and April and probably close to 2 million infections. Testing availability has improved a lot since then, but it's still not adequate to catch every case.
It has gotten easier, though, to make educated guesses about the true extent of Covid-19's spread by combining what has been learned from those antibody surveys with metrics such as the extent of testing relative to population and the percentage of tests coming back positive.
Two websites that now make these estimates on a daily basis are covidestim.org, the joint product of researchers at the Harvard and Yale schools of public health, and covid19-projections.com, the work of New York data scientist Youyang Gu. The latter decided in September that his much-lauded forecasts of Covid cases and deaths were no longer needed amid a surfeit of new forecasting models, but revived and updated his infections tracker late last month. Here's how these two infections estimates compare with the confirmed-case count.
Covid19-projections publish its estimates with a two-week lag, which is why the gray line in the chart stops where it does. As of that stopping point, which was November 15 when I downloaded the data Monday afternoon, the two models delivered remarkably similar estimates of the cumulative number of US infections: 52.6 million according to covid19-projections, 52.2 million according to covidestim.
Those both come to just under 16 per cent of the US population. As of last weekend the covidestim infections estimate had risen to about 18 per cent of the population; the covid19-projections estimate will likely come in a little bit higher than that. That's well short of the 70 per cent that may be required to achieve the herd immunity that will cause the pandemic to fade on its own. But in conjunction with social distancing, mask-wearing and other changes in behavior it can help slow the spread. And in conjunction with the vaccines that should start to become available later this month it could conceivably end the pandemic soon.
These estimates thus contain useful information that ought to feature more prominently in media reporting and political discussions about the pandemic and how to end it. And while the above chart makes clear that the two models are in far less agreement on the exact trajectory of the pandemic than the cumulative infection totals, averaging their results together - as the Centers for Disease Control and Prevention does with Covid-19 forecasts - delivers a possibly enlightening alternative history of the past nine months.
Here, for example, are population-adjusted infection estimates for the states hit hardest by the pandemic last spring (New Jersey and New York) and this fall (North and South Dakota). Yes, the Dakotas epidemic has been bad, but not nearly as fast-moving as the explosion of the disease in and around New York City in March. If these averaged estimates are anywhere close to right — and Thanksgiving gatherings don't cause a resurgence — it also appears to have topped out at lower levels of prevalence. Which makes sense: Despite the much-publicized reticence of the governors of North and South Dakota to take aggressive measures, lots of people in the Dakotas have been working from home and otherwise social distancing in ways that almost nobody was doing in New York in early March.
Now here are the five most populous states other than New York (where the spring wave was so big it would be hard to differentiate the other states on the chart if I included it):
Covid outbreaks in California, Florida and Texas have gotten tons of attention over the course of the year, in part just because the numbers involved are so big and in part because they have been enlisted to advance various political narratives. But Illinois and Pennsylvania now have higher prevalence of the disease than either Florida or Texas during past peaks, while California has been a standout in keeping Covid-19 under control - although it does appear to be having its worst-yet outbreak now. (Bloomberg)