re: How COVID cases are defined and counted by the CDC - this will piss you off

quote:

---

and, as suspected, I see LA is not separating them out.

And we are to believe the numbers?

---

Well, testing has some interesting setbacks.
So, for instance, the PCR testing sensitivity has been reported by my local pathologist to be in the 70-85% sensitivity range.
English: out of 20 people who definitely have the virus, the test will show 14-17 positives.

When I have someone in my office that got sick 1 day after being in contact with a household contact who tested positive, I know that testing them and only counting the ones who test positive will artificially lower those numbers. It also doesn't change what needs to be done for the patient in front of me.

It's an inherent flaw in epidemiology and public health:
1) test only will result in wrong numbers. In the setting of PCR tests, you can generally assume no false positives, but false negatives (and with the test in my area, a pretty good false negative rate). In the setting of other types of tests, you can increase the sensitivity (fewer false negatives) but you're going to start trickling in false positives
2) suspicion only will give a different false number. Could be high, could be low. Could be pretty close to accurate. Depends on actual amount of disease present, how aware the clinical community is, and how easily identifiable the clinical illness is. So, for chicken pox, this is appropriate and pretty accurate. For new disease for which no unique symptoms or clinical identifiers exist, it's probably not the best.

There are special codes I use for diagnosis based on a positive test (u07.1) vs suspicion due to close contact (z20.828). There either is or is talked about another emergency "u" code "u07.2" (I think) that is either talked about and not integrated into my coding software, talked about and never implemented, or not yet deployed meant to be used more or less for when a doc is "pretty sure" you have it.

The problem with all of this is that it doesn't actually effect the doctor-patient level of care. That is a super fragmented system that gets reported on paper or in 100+ electronic forms that have to all be able to get data to the epidemiologist for their use.
The problem is that is a slow process, the epidemiologists/public health experts know this, so they really do understand that they aren't working with 100% accurate numbers but do their best to come up with an accurate estimate. They will essentially always be wrong. But they have to try to get decent estimates, and that is usually going to involve "hard positives" from testing in conjunction with "best clinical guess" to try to get a more representative number.


How nefarious those in power are with the estimates is another issue entirely, but the data collection itself will always be less than 100% accurate from a single source, and I don't think that anyone who actually collects and uses the data would pretend it was. But the estimates will improve over time and, ideally, we have a benign and benevolent enforcer of policy based on bad data or, at the very least, understand where the shortcomings are.