I wanted to write a bit about why we can’t rely on all this data too heavily, and the fog is getting worse.

First, what do we really want to know?  I can think of several things I’d like:

MeasureWhy it mattersHow to measure itWhy we can’t right now
How many/what percentage of people in king county are infected?This would tell us how we’re doing – we’d see the shape of the curve and know when we’ve actually flattened the curve and (essentially) eliminated the virus in King county.Do periodic random sampled testing of the community.Not enough tests
How many/what percentage of people in king county have been infected?This one will be crucial to understanding rates – death rates, hospitalization rates, etc.  It also will tell us how close we are to getting herd immunity.Do Serum tests of a random sample of the community.Not enough tests
How many people in King County have died of Coronavirus?


Important to know how deadly the virus is, and to which populations it’s most deadly.  This could be crucial for opening up safely.Test everyone who dies of suspicious causes to confirm the CV cause.Not enough tests

If we had these three, we’d have an excellent picture of what’s going on.  I’d argue we have a decent approximation for 3.  But the first two are inconceivable; you’d need maybe 1000 extra tests a week, and it’s unthinkable to divert these precious tests from patients and healthcare workers.  We can’t measure them well.  We need more test kits!

To see why let’s talk about the metrics we have:

MetricWhat it’s a proxy forWhy it’s okWhy it’s not
Number of positive tests.Number of infectionsOver time, as infections grow this should increaseNot randomly sampled (and the criteria for testing has been changing), and number of total tests has been growing (which would make infections appear to be growing).
% of tests that are positiveInfection rateTells us something about incidenceAgain, not random sampled and the criteria keep changing.
Number of deaths tested positive for CVCV Deaths, CV Death ratePretty definitive dataLimited tests mean we may be missing some folks who died but didn’t get tested.

So, with this all said, let’s talk about some things that I’m pretty certain are happening.

  • First, it seems certain we’re missing a lot of infections, especially ones with mild or no symptoms. Nobody is going to fight to get tested because of a fever and cough that resolves in a few days. We almost certainly have a lot more infections than we know about – but have no way to know how many there are.
  • Second, part of the growth in our positive tests results from a growth in tests overall. However, our testing has really leveled off recently so I doubt this has been an issue for the last week or so.
  • Third, I suspect our number of infections in King County may be dropping, but the measures we have are now insensitive to this. To understand why, imagine this situation.  2 weeks ago, we had 10,000 people with COVID.  They came in and begged to be tested, but we only had 1000 tests to give them.  We’d test 1000 and get 100 positives (because only 10% of people with symptoms have COVID).  Now it’s a week later and we’ve been successful: we have 5000 people with COVID.  We do our 1000 tests and now and we still get 100 positives!  Because the incidence of COVID so outstrips our ability to test, we’ll always be able to find COVID patients as long as there are enough sick people showing up at the ER!  Of course, at the limit this will end; when we only have 50 COVID patients the tests will start to drop.
  • Fourth, there are lags everywhere. People get infected, 5 days pass before symptoms. Then they take 7-10 days to get sick enough to show up at the ER.  Then they get tested, but the test results take 6-8 days to come back.  It could take 23 days before social distancing shows up!  And it may take 30 days for you to die. So anything we do will take 2-4 weeks to show up in the data.

It’s a really hard set of problems.  The bottom line is we work with the data we have and try to have faith in the basic fact that if we don’t get near to each other, we can’t get the illness.

And, we should be grateful that even with all this fog, the data still shows that our efforts are working.