Do CDC's "COVID Community Levels" account for home testing?

On the assumption that most people who test positive for COVID using a home testing kit don’t report their condition to local health officials, it seems like a no-brainer to say that the official statistics are undercounting the actual incidence rates for the disease. The CDC’s COVID community level scheme defines its levels based on official incidence rates as well as hospital utilization metrics: for an incidence rate of less than 200 cases per 100K people per 7 days, the level could be low/medium/high depending on hospital utilization, but for an incidence rate above 200, the level is either medium or high.

So was that cut point of 200 chosen with the advent of home testing in mind? Are there estimates out there of how many positive tests (due to home testing) aren’t being captured by the official statistics?

Interesting question. I’m pretty sure I’ve had influenza several times in my lifetime, yet I’ve never sought medical attention for those events.

There are no credible estimates of infections detected from at-home lateral flow (antigen) tests. It isn’t clear whether CDC has accounted for this in any kind of statistical sample analysis for the CDC community level scheme but aside from voluntary tracking and reporting per smartphone applications like the ZOE COVID Symptom Study App the only real way to get any accurate statistical count on infections and the spread of particular variants is to look at hospitalizations. Unfortunately, hospitalizations are a lagging indicator of spread and any trends you see from infections found in hospital are lagging contagion trends in the general population by a minimum of 3-4 weeks.


That assumption is shaky UNLESS most people who test negative on a home test report their results.

I would wager that even fewer people report a negative result than report a positive result. But we’ll never know, because (by definition) such things are not tracked. It very well could be that the official reports are undercounting the per-capita incidence. But there simply is no evidence either way – with the possible exception of waste-water testing.

Wastewater surveillance allows tracking of the presence of different strains of SARS-CoV-2 but provides only qualitative data about incidence and timing. It does not indicate how many people are infected, and is also a lagging indicator because you have to have some threshold of infected people before it will statistically be observable. It’s a good tool but not one that provides quantitative information the way individual testing does.


I’m not understanding this. Since the stats are reporting the total number of new cases each day, why does anybody need to report negative home test results? a report of 5 positives and 10,000 negatives from home testing just gets listed as “5 new cases today”; so does a report of 5 positives and 50,000 negatives from home testing.

I’m not sure what you’re saying. What you usually see in graphic form are daily positive cases.

My county also reports total tests (positive and negative) and the positivity rate, but that’s not what’s on the front page. Positive tests are slowly inching up but total number of tests have remained low. Therefore, our positivity rate has risen from a low on April 5 of 1.2% to 17.7% June 15. Part of that jump in positivity is due to about one-third less tests being reported. The other part is that there are about 8X more daily positive cases.

I thought wastewater surveillance was a sort of canary in a coalmine method. Levels start rising before you start to see a rise in cases because people start shedding virus before they get symptoms or if they never get symptoms.

It certainly serves as an indicator that a mass outbreak is occurring, especially in lieu of adequate sample testing, but it isn’t as if you’re going to find detectable levels of viral RNA (which is what it is looking for) in the sewage waste stream if only a handful of people are infected. What it does really well is give an indication of the rate of change of community spread and what variants are dominating, probably even better than individual testing because regardless of whether a variant is producing more or less symptoms or hospitalized cases, it will appear in relative proportion in the sampling of the wastewater stream. The blue highlighted box in your link spells out the advantages of wastewater surveillance. It is obviously not a replacement for public testing programs but it gives what is essentially a completely unbiased sample of the distribution in the community regardless of how well it is served by healthcare or or testing availability.


Ah. Okay. I think we are in agreement.

It should be very easy to understand. You cannot know the positivity rate without knowing the total number of tests, both positive and negative. And without knowing the positivity rate, you cannot get the full picture of the spread and intensity of Covid.

This is true, but this is the first time in this thread that you have mentioned positivity.

My initial question wasn’t about positivity rates (which do indeed require knowing the total # of tests). I asked about incidence rates, i.e. the total number of new cases each day.

My assumption was that most people who test positive for COVID using a home testing kit don’t report their condition to local health officials. In response, you said:

which doesn’t make sense. Negative test results are irrelevant to calculating incidence.

This is why I’m looking at hospitalizations. If hospitals test everyone who comes in regardless of reason, it’s a sampling of the population (if somewhat biased).

The only thing test positivity rates are telling you is how many people who are symptomatic or were in a situation where they were exposed to someone who was infected and and presumably shedding have actually contracted the virus and even that is pretty suspect because the lateral flow tests will often show negative a day or two into symptom onset. It isn’t an unbiased measurement and without consistent sample testing it is not really a great leading indicator of outbreak. It’s a piece of the puzzle but not the whole, and as you indicate the real measure of interest are the total number of new cases, and of course hospitalizations.

It’s actually really biased because people don’t generally go to the hospital unless they are ill or are having some kind of procedure. The young and nominally health segment of the population who are the most likely to asymptomatically spread contagion are the least likely to go into the hospital. Intake testing data isn’t worthless as it does give an indication of spread in the portion of the population most likely to be impacted by the virus but it isn’t representative of the population as a whole.


Many places, including where I live, do regular sewage testing. That’s probably more accurate and certainly more timely than counting hospitalizations.

Our local numbers are quantitative. I believe the average amount of virions per infected person shed varies by variant, but within any wave the curve of sewage data has tied very well to every other indication, and is a leading indicator.

Where I live, there are still a lot of people being tested prior to medical procedures and travel (although the demand for travel tests is about to plummet) so the positivity rate is somewhat meaningful. And yeah, underlying the “positivity of symptomatic people” is how much of the respiratory bugs that are making people sick enough to seek testing are actually covid. That being said, I’ve stopped paying attention to both case counts and positivity, and am only paying attention to wastewater and how many of my friends and other contacts have covid. (there’s still a great deal of chatter about that in my circles.)

The wastewater is certainly as unbiased of a sampling method as you could practically hope for without going out to peoples’ homes and subjecting them to mandatory sample testing, but it does require a certain threshold to get to the point of detectable level of virus. To be clear, I’m not objecting to it—it’s a great tool for understanding community presence of the virus and also tracking the evolution and spread of particular variants—and while wastewater surveillance has been used before in underdeveloped nations where sample testing of the general public is impractical to look for outbreaks of contagion like cholera or typhoid, this is the first time it has really been used in this way as a means of surveilling the spread of a respiratory virus. But it does have limits in terms of the necessary threshold for detection and can obviously not be used for any kind of test & trace effort.


It hasn’t been zero locally since we started using it. I don’t think the threshold for detection is all that low.

We’re still well into a pandemic (even though many people are still vigorously denying that there is any concern whatsoever) so community levels are not negligible. It is also the case that many people will continue to shed detectable RNA even well after the virus is no longer active in their system, so there is a long ‘tail’ of residual detectability in rtPRC resting that I assume also occurs in wastewater surveillance.


That brings me to another thought; does society care much about COVID if it’s no longer putting people in the hospital and/or killing them at the rate it was before? In fact, according to this Financial Times article, Omicron is LESS deadly than influenza., per 100,000 infections. Of course, the caveat is that the amazing transmissibility of Omicron vs. influenza means a lot more people will be infected.

It does bring up an interesting thought experiment; what would have happened if instead of COVID in late 2019, it was an extremely transmissible variant of influenza that came around the world?

Even if all test results were reported, that wouldn’t tell you the incidence rate, because some people will have it but won’t be tested: Remember, covid has a very high asymptomatic rate, and even some folks with symptoms might ignore them, or decide that they didn’t have anywhere they need to go in the next couple of weeks anyway, or whatever.

But what could tell you the incidence rate is testing of a sub-population that’s equally likely to be infected as the total population. Ideally, this would be a truly random sampling of the population, but that’s really hard to do. But there are circumstances under which people need to test regardless of their symptoms or lack thereof: Tests might be required for travel, or a workplace might require periodic tests, or the like. And those are probably a relatively representative segment of the population, and the results of those tests are likely to be reported to public health authorities either way.

I’d be really careful about accepting those assertions without some strong caveats. Part of the reason that the ‘Omicron’ variant is producing less mortality in populations previously exposed to previous variants is because the original ‘wild-type’, ‘Alpha’, and ‘Delta’ variants have already clearcut through the most vulnerable populations, and many people have at least some degree of immunogenicity through prior exposure and/or vaccination, so even though ‘Omicron’ is more transmissible and is infecting vaccinated people, the severe morbidity and mortality is less. In naive populations such as mainland China and Taiwan where they effectively suppressed earlier variants the observed virulence of ‘Omicron’ is not nearly as benign.

This is the scenario that epidemiologists have been warning about for decades and it could still happen. We have existing antigen tests for the common strains of Influenza A, existing approved vaccines that are at least marginally effective in preventing severe illness and death, and some therapeutic treatments (although most care is still palliative) but a really aggressive avian or swine flu could be devastating. SARS-CoV-2 is a love tap of a pandemic but if it ever developed the virulence of other betacoronaviruses like the original SARS-CoV(-1) or MERS-Cov it could deliver a truly massive blow to population health, social institutions, and the stability of the global economic system.

The people who watched the 2011 film Contagion and scoffed at how unlikely that is should realize that if anything it actually understated the effects of a pathogenic virus with a >1% infection fatality rate, and the only thing really unrealistic about the film is how quickly a vaccine was developed and how universally vaccine uptake was embraced; virtually everything else in the film, from government infighting and suppressing information, economic and social disruption, people using social media to spread information and profit from scams, and officials using inside information for their own personal profit and interest, has come so exactly to pass in the comparatively benign COVID-19 pandemic that calling the film prophetic is an understatement (although not surprising given how much research and composites of characters were drawn from epidemiological history and specifically the 2003-4 SARS outbreak that nearly became a pandemic).