Testing early, quickly, accurately, and thoroughly is important in the war against COVID-19.  Not only does it allow us to identify those who have been infected early in their infection, it allows us to track all their contacts more thoroughly, it allows us to treat them more effectively, and it also allows us to restrict their travel and isolate them before they have a chance to infect others (the 4T program to fight epidemics).  All 4 steps are necessary to allow a country to effectively control its epidemic at minimum cost to society.  Countries that have effectively adopted this program such as South Korea and Singapore have been able to control their epidemic without resorting to Draconian measures such as those employed by China who quarantined entire cities and provinces because they didn’t know the full dimensions of their problem.  

On this score, the US has done very poorly — starting slowly, with numerous setbacks and a far too restrictive set of test protocols.  Testing was impeded in February by several CDC errors including the requirement that they verify all testing results centrally.  Testing improved significantly after the CDC removed this bottleneck toward the end of February and permitted private testing labs to do testing.  Since then the number of tests completed (top line in the graph below) has grown exponentially to keep up with the exponential growth in infection (positive results in the graph below).  A good sign but not ideal. 

Ideally, you want the number of tests performed to be significantly greater than the number of positive cases that you suspect is in the population to make sure that you are counting every infected person.  As of yesterday, 3/28, the US has tested just 0.2% of its population.  By comparison South Korea reached this milestone on 3/1 and has now tested 0.8% of its population.  2.4% of its tests came back positive while 15.9% of US tests came back positive.  Unless we have good reason to believe that our underlying infection rate is really 15.9% we could still be under testing by a factor of 6.5.  So we are doing better but not good enough.

Under-testing is a critical problem because unless we have an accurate count of the true number of infections we may be over-estimating the mortality rate or deadliness of COVID-19.  In February, the measured coincident mortality rate {deaths (see the fourth line in the graph above) divided by positive cases} was as high as 8% on 3/3 because the denominator was biased too low.  As we tested more the denominator grew faster than the numerator and this coincident mortality rate fell to near 1%.  This, however, was biased too low.  As we had pointed out before, it takes -1 to 21 days for diagnosed cases to result in death or cure (-1 for many cases diagnosed after death), with most taking 3–7 days with a median of 5 days yielding our lagged5 estimate of 4.5% for the US.

Not every state has tested with the same degree of rigor.  The table below ranks the states by its hotness measure — i.e., the number of confirmed cases per million population.  The testing metric in the last column show how thoroughly each state has tested its population and how likely or unlikely its results are biased.  On this score NY and WA have done the best and NY mortality rates are least likely to be biased.  WA estimates are probably biased too high because of the source of their infection in a nursing home bias the age of their dead too high to be representative of the state.  On the other end, California had some of the earliest cases in the country but their testing has lagged. They have a huge number of pending cases that should be resolved soon and improve the accuracy of their numbers but even then they would still be much worse than NY and WA.  Our guess is that California’s confirmed case count is way too low. 

One final note, the third line in the graph above shows the number of cases serious enough to require hospitalization.  Many of these cases will end up in the death count whether any drastic action is taken tomorrow or not.  They serve as a leading indicator of future problems if they soar out of control and overwhelm the capacity of our hospitals.  That they are growing exponentially is not a good sign.  

InfectionsDeathsMortality CoincidentMortality
NY           2,745533938831.7%3.3%0.8%
NJ           1,252111241401.3%3.6%0.3%
LA              71333151374.1%9.1%0.5%
MA              6134257441.0%3.3%0.5%
WA              56643101894.4%7.6%0.8%
MI              46646501112.4%6.0%0.1%
CT              4271524332.2%5.2%0.3%
CO                3582061442.1%4.8%0.2%
IL              2753491471.3%2.8%0.2%
FL              2424038561.4%3.4%0.2%
GA              2232366692.9%6.3%0.1%
PA              2152751341.2%3.6%0.2%
CA              14055491192.1%4.5%0.1%