Top: before noon on 3/16, bottom: after noon on 3/16.  Data from https://www.cdc.gov/coronavirus/2019-ncov/cases-in-us.html

Same page after 3/17 update And after 3/18 update

I was waiting with bated breath for the CDC update on cases of COVID-19 in the U.S. by date of onset. To my eyes this looks like absolutely wonderful news - that it means that the exponential growth of the virus has been "flattened" - but I'm still afraid I may be misinterpreting what I see. It bothers me that the headline for the graph hasn't changed. But it APPEARS that the number of new cases stopped increasing after 3/3. Am I missing something?

Top: before noon on 3/16, bottom: after noon on 3/16. Data from https://www.cdc.gov/coronavirus/2019-ncov/cases-in-us.html

Note: In today's update (the first in three days) the CDC page shows the number of cases with onset on 3/4 to be 159, and 3/9 is 307. This doubled in five days rather than every two or three as other numbers would suggest, but given the dramatic changes in this curve, and their overall direction, I think the top answer below has been vindicated.

  • 6
    $\begingroup$ Did you read the text in the gray shaded area? $\endgroup$
    – Bryan Krause
    Mar 16, 2020 at 16:54
  • 1
    $\begingroup$ The US has been adding more than 500 new cases a day over the last couple days; these data seem very incomplete. I would wait to have more information before interpreting them much. (There are over 4000 US cases but these data are based only on fewer than 800) $\endgroup$
    – Bryan Krause
    Mar 16, 2020 at 17:03
  • 2
    $\begingroup$ This is against the experience of all other countries dealing with this disease. So this is most like due to under-testing. $\endgroup$
    – Chris
    Mar 16, 2020 at 17:40
  • 2
    $\begingroup$ Perhaps worth noting that the CDC is not keeping its public data current. $\endgroup$ Mar 16, 2020 at 18:13
  • 2
    $\begingroup$ Any data you get from the CDC will not be current and so not accurately reflect trends. Further, lack of testing in the US means that any data that you do see will not accurately reflect true infection numbers or rates. $\endgroup$ Mar 17, 2020 at 20:03

1 Answer 1


The information you show is meaningless unless it is also accompanied by the number of tests performed. If the CDC tested 2 people and got two positive results, that doesn't tell us anything at all about the actual rate of infection. And, indeed, the CDC has been testing a shockingly low number of cases:

number of tests performed by the CDC vs public health labs

‡ Data during this period are incomplete because of the lag in time between when specimens are accessioned, testing is performed, and results are reported.

(image source)

The last day they give data for, 2020-03-15, includes 0 samples from the CDC and 4 samples from other public health labs. Since only 4 samples are represented in the chart, of course we see a drop in cases, but that's because they are based on only 4 tests. As more tests for those dates and subsequent dates are reported, the numbers will change.

Date Collected  CDC Labs    US Public Health Labs
3/12            75‡         2302‡
3/13            0‡          1511‡
3/14            0‡          165‡
3/15            0‡          4‡

As you can see in the table above (from the same source as the graph), the past few days have seen almost no tests are being included in this analysis either from the CDC or elsewhere, so any drop in cases is completely meaningless.

  • $\begingroup$ The dates you highlight - from 3/12 on - are within the gray box, which we already know is incomplete data. The days outside the gray box are the ones more at issue. I see today's update increased the total cases in the chart from 930 to 1295. The number of cases on 3/9 was 113, but all the others were less than the peak of now 90 (not 70 any more) for 3/4. There is something complicated going on here and I think your connection to this other page IS on the right track but it is wrong to say no testing is happening or to link it with the gray box. $\endgroup$ Mar 17, 2020 at 16:30
  • 1
    $\begingroup$ @mike what day are you referring to? The truth is, sadly, that the answer won't change. There is no way of looking at these data that could make them be representative of the real infection levels in the general populace, but if you clarify exactly what dip you are referring to, maybe we get get some more detail. – terdon 43 mins ago $\endgroup$
    – terdon
    Mar 17, 2020 at 22:10
  • $\begingroup$ Sorry to be confusing - I've gone back and edited the original question to add the third (current) version I was referring to. Note that there have been cases added, but they have been added throughout the curve - so the 3/4 peak increased, yet 3/5 to 3/8 are still less than that. So while the reported cases have been increasing exponentially, those at least might be OLD reported cases. That means the total infected is much higher than we think ... but also leaves open the question of whether recent changes have actually reduced the spread of the disease. $\endgroup$ Mar 18, 2020 at 12:43
  • 2
    $\begingroup$ @MikeSerfas any effect of the recent changes will only begin to be visible in a few weeks, not now. The main point, however, is that those data are not really helpful if you are trying to monitor the actual rate of infection because the tests being reported are way, way too few to accurately represent the rate of infection in the population. And, again, the dip isn't informative since there was a corresponding decrease in the number of tests whose results are being reported. $\endgroup$
    – terdon
    Mar 18, 2020 at 13:59
  • 1
    $\begingroup$ @MikeSerfas you're still missing my point. You keep looking at the graph and ignoring the big grey warning on the last days explaining that "Data during this period are incomplete because of the lag in time between when specimens are accessioned, testing is performed, and results are reported." And you also don't seem to be taking into account the number of tests performed. Wait a few days and you will see the numbers for this period will change significantly. I understand the wish to see this flattening but there simply is no data to support that. $\endgroup$
    – terdon
    Mar 19, 2020 at 9:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.