Covid numbers driven by increased test coverage?

Great Debates
1

I tried to depoliticize this enough for QZ and I realized it was hopeless, so it will be a debate.

In popular awareness, we’re judging our COVID-fighting performance by the metric of new cases per day. Depending on which geographic area you measure, this metric seems to be rising in many places. Some people are saying this is inaccurate and I want to explore one of the theories.

Not to bury the lede here, Trump has made several public comments to the effect that we’re getting so many cases because we’re doing more testing. He’s a stupid man and a terrible speaker, so it’s easy to believe that Trump means to say “testing causes COVID”. That’s obviously a ridiculous belief and there’s no point in discussing it. I’m also uninterested in whether Trump believes that or not.

Instead I want to explore the question that hides in the more charitable framing: do the rising numbers reflect actual COVID cases, or do they reflect something else?

I’m going off the numbers in the NY times COVID graphic.

Now… we know that testing took a long time to ramp up. It seems sensible to assume there were (and are) far more cases out there than we knew about. If the cases we tested are the tip of the iceberg, is it not reasonable to believe that as we administer more tests, we’ll get more positive results even if the real case number is declining? It seems like that’s in the universe of possibilities.

If that were true, then we’d see the case rate fluctuating, but the death rate would continue to march steadily downward. And that’s what we’re seeing. The case rate in the US spiked and flattened, though it’s fluctuating in various areas. But deaths overall have been falling steadily falling since April 15th!

Another widely held idea is that red states are having a “coronavirus explosion” due to overeager relaxation of rules. It’s true that there seems to be a correlation here, but correlation is not causation. Is it also possible that we’re getting more test coverage in regions that had poor test coverage, at the same time that small case increases in low-populated counties is causing a rapid apparent doubling?

I don’t have any answers here, but it seems like there is support for the idea that because test coverage itself is changing rapidly, the reported case load is not reflective of actual incidence of COVID. Anyone care to set that straight?

2

I prefer this site as overlays the graphs from each state:

Taking California as an example - It shows the 1 wk average daily new cases new cases increasing rapidly (but linearly, not exponentially). Switching to 1 wk average daily new tests, the slope of the graph is about the same. In contrast, 1 wk average daily new hospitalizations and deaths remains more or less flat. One would assume hospitalizations would correlate to number of cases but they don’t appear to. Ergo, one could assume the increase in cases is due to the increase in testing.

3

It’s certainly plausible.

4

I’m a bit confused by the OP premise.

If we were simply counting the total number of cases found while testing, well, that number will always be going up in an active pandemic. The more we test the more rapidly it will grow. But we’re talking about rates of cases found per 1,000,000. Isn’t that what we’re talking about when it comes to reporting trends of whether the spread of the virus is growing or diminishing?

5

If the number of cases is always going up, and a million is always staying constant, then the number of cases per million will also always be going up.

6

Well yes, I think that is the question the OP is asking, whether it does any good to being screaming about the “infected” rate going up, or whether or not the rate of infected is actually going down …

7

I’m not talking about that. I’m talking about new reported cases per day.

I am asking how confident we are that reported cases per day represents actual new cases per day in the US, given that:

  1. We know reported deaths per day are consistently and linearly decreasing
  2. We know that testing coverage started at a nonexistent level and stayed that way for a long time
  3. In the areas of the US where COVID is “rapidly doubling”, in many areas we’re talking about single digits in places that had no existing caseload.
8

Testing is increasing in NY and IL and yet the case numbers aren’t increasing, and these are the two states still in restrictive lock down.

9

It seems to me that this is an easy enough question to answer by comparing cases to hospitalizations and deaths. Here in Ohio the health director said that those metrics lag 2 and 3 weeks behind positive case numbers, respectively, so to some degree the answer is “we’ll see soon enough.”

But in a nutshell, the question is whether or not we’re now finding positive test results in asymptomatic or mildly symptomatic people who we were unable to test 2-3 months ago due to lack of supplies. And as far as I can see, the answer is no – here’s the ODH dashboard for CV19 trends, and cases have not diverged noticeably from hospitalizations and deaths. Perhaps other states have similar metrics.

10

Compare and contrast with Arizona, however; the one-week average daily new tests is a shallow upward line, whereas the one-week average daily new cases is going up more rapidly. They’re doing more tests, but more of those tests are positive as well. (One-week average test positivity had been in the 5-8 percent range for some weeks, but crossed 10 percent twelve days ago and is now approaching 18%.) California’s test positivity rate is near flat, but Florida’s, to name another, is increasing. When you are doing more tests AND a greater percentage of the tests are positive, it is likely that actual incidence is rising as well.

11

Here in Kansasits all over the map. They finally got around to testing some rural counties and found thousands of people infected.

Lately though the numbers have begun to flatten with occasional spikes. In my area they test about 200 a day (sometimes 1,000 or more) and are finding about 1%-3%. So the total number of cases still goes up.

Now one question I have. What is the reliability of the test? Since number of positives is around 3% isnt that about the rate where the tests could be right or wrong?

Very confusing.

12

The one metric that is probably a near-all value (in the US) is the fatality count. From that, you should be able to back-calculate the total number of infected in that region, using statistics from South Korea and making the assumption that South Koreans and Americans are both just humans and Corona is just Corona, wherever you go (plus or minus a bit). The only real disadvantage with that is that the fatality count will be a small enough number in some places that it’s too noisy to deal with. In situations like that, you can group together a few regions that are low-pop and consider them to be like a single state.

Fatalities occur about 11-12 days after infection. If things just spiked, you won’t see a matching result in fatalities for a week and a half, so you need to wait at least that long to determine whether there’s a change in test taking or a change in infections.

13

The information I’ve seen suggests fatalities occur on average about 18 days after onset of symptoms, which itself occurs 5 to 7 days after infection. Have you seen something more current?

14

My memory is that the metric was from testing positive to death (as opposed to “from exposure to the disease to death”) but, I believe, that I got it from CDC data and near the beginning of April (the same time as your article was written) seems like it’s probably about the same time frame.

I can’t track it down at the moment and you have a cite, so I’d suggest to others that they go with your number.

15

People keep pointing this out, so I want to clarify that I understand this. The 7-day average of new cases per day slowly drifted downward to about 24,000 and has stayed there for the past 5 weeks. By contrast, the 7-day average of fatalities has trended downward almost every single day and is on a trend to be in double digits, nationwide, by late summer.

Granted things could look very different in 2 more weeks… in my state there seems to be a very convincing correlation between bad governmental decisions and spikes in the virus. But the national fatality rate keeps marching downward independent of all these other metrics.

16

No, it’s not on that trend, because the nationwide trend conceals a lot of local variation. In particular, much of the downward trend has been concentrated in places like New York, which got hit hard early; their 7-day average has fallen from 1200+ to 65 over the past two months. In other states, the trend line is at best flat or slowly rising. See, e.g., Arizona, where the 7-day average of fatalities has risen from 12 three weeks ago to 21 now. In Texas the death rate is down from its April highs, but it’s still bouncing around in the same range as it was a month ago, at about 25. Tennessee is bubbling along at a fairly steady 8. Nebraska’s rate has never been high, but it is slowly increasing. California’s 7-day average has been bouncing between 60 and 75 for the past month; it’s currently 65.

If California, Arizona, Florida, and Texas continue their recent trends and every other state drops to 0 deaths, we’re still going to be in triple digits through the summer.