In the current issue of the Carmel Pine Cone, there is an editorial that makes this point:
But one thing we’ve been warned about from the very beginning of this crisis, and which obviously still holds true, is that test results and case rates are inherently unreliable ways to measure the seriousness of an epidemic because both numbers are heavily influenced by human choices, if not by politics. While nobody goes into the hospital or dies to make a point, you
could lower this county’s positivity rate or case rate simply by focusing testing efforts in places where the virus isn’t very prevalent — which Monterey County could do by testing mostly in the Peninsula. Or you could artificially raise either metric by testing where the outbreak is most severe, which this county could do by mostly testing in Salinas and the Salinas Valley. Which is exactly what we’re doing.
Considering what’s at stake (the economy), it’s imperative for the governor to get it right. He’s right to use current numbers. But instead of shutting everything down based on case rates and positivity rates, why isn’t he doing it by hospitalizations and fatalities?
There are several problems with what the Pine Cone editor is proposing. The most obvious one is timing, or currency. If you’re going to be making decisions about what to do in a pandemic, you don’t want to rely on lagging indicators, or you’ll always be behind the curve. Anyone who has studied control theory knows that delay is the enemy of stability. Relying on lagging indicators is a recipe for oscillations, and oscillations in the rules-setting process will not only make the rules less effective, but will also make the public less likely to follow whatever rules are in place. The average lag time from infection to hospitalization is on the order of 2 weeks. Add another week or two to that for the average lag from infection to death.
The point about gaming the testing results is accurate, but the solution is more specificity in how the results are tabulated, not in throwing the testing baby out with the bathwater. There are several problems with the way that the positivity numbers are calculated in California. The first is that both the denominator and the numerator relate to the number of tests performed, not the number of individuals tested. The second is that purely voluntary — “Gee, I wonder if I have Covid-19?” — results are mixed in to the samples used to calculate the numbers. So are screening tests on older people in institutional settings. Ditto for health care worker screenings. All of those things are important, but the criteria for reopening should be dependent on how well the public health system is controlling the epidemic, and in the absence of widespread random testing — and we’re a million miles from that — the testing that matters is the testing performed at patient-intake settings and during contact tracing. We should be keeping track of the number of positives in that subset of the testing, and of the positivity ratio.
I have always though it should go without saying that antibody tests, which determine if the subject of the test has ever been exposed to the virus, should never be mixed with molecular tests that determine if the subject is infectious. From a public health perspective, you want to identify and isolate the infections patients. In the absence of any information about susceptibility to reinfection, counting people who probably have had an infection at some point in the past is not relevant. If you want to keep track of that for looking at herd immunity or trying to find the ratio of asymptomatic, infected people, fine, but do it separately.
OK, why is the positivity ratio (number of positive individuals over number of people tested) important? Because it tells you how well you’re doing with your contract tracing, and control of the disease depends on tracking down and isolating infected people, many of whom are asymptomatic. WHO estimates that contact tracing positivity rations need to be under 5% to have the disease under reasonable control.
Which brings me to another important point about contact tracing. It only works well when the test turnaround time is short. The infectious period of the disease appears to be about six days. If you test a person at the beginning of the infection and your test turnaround time is six days, that means that by the time you get the result, the patient is no longer infectious. If the person is not isolated until the test results are received, people were being exposed during the entire course of the disease, and you’re locking the barn door after the horse has disappeared. This is not an academic point; in Monterey County, test turnaround time is averaging 3 days.
While I’m on the subject of testing, there’s another problem with the current molecular polymerase chain reaction tests. They report as positive patients who are not infectious. In a sense, they are to good for the purpose for which they are being used. Part of their over-sensitivity is useful: on the front end of the disease, they can detect patients who are about to be infectious, and they can be isolated to limit the number of people infected. But the window is only a day or so. On the back end, a person can test positive in a polymerase chain reaction test weeks after they have ceased to be infectious. We really should be using less-sensitive tests with faster turnaround.
Another point is that deaths and hospitalizations are not good indicators of the relative prevalence of the virus in the community if the demographics are not held constant. Let’s say that a cohort of young people is experiencing a rise in infections. Those people are far less likely to be hospitalized for or to die from the disease than older people, so looking at deaths and hospitalizations would allow the disease to spread widely in a community whose infection profile tilted young before those metrics finally started to spike as the young people secondarily infected the older ones.
In a nutshell, the problems with what we’re doing now are:
Even with the imperfect measures that we have, we can tell that we’re not testing enough. In fact, the rate at which we’re testing is dropping.
And the way we can tell that we’re not testing enough is that the positivity rate is too high:
This is a local problem. Note that the Monterey County positivity ratio is over 12.5%, and has been hovering around 15% for some time, while the statewide positivity ratio has declined from just under 10% to about 4%.
Our too-high positivity ratio surely means that we are undercounting infections.
And we shouldn’t change our criteria for opening to deaths and hospitalizations. When you’re controlling an epidemic, you need up-to-date knowledge, and the lag for those is too long.