Talk:COVID-19/Adjusted Number of Cases
Add topicLatest comment: 4 years ago by Dan Polansky in topic Base of the confirmed case growth vs. base of true case growth
Adjusted Number of Cases
[edit source]- You seem to propose that the average geometric growth over the past seven days is applied to "today" to get a prediction of the count in seven days' time, for exponential growth. I don't understand why you call this an estimate for "today". For example, if N = 2^d, where d is the number of days, and today is d=8, then N(1) = 2 and N(8)=256; so your "estimate for today" is N(15)= 256 * 256/2 = 2^15 = 32768. But really that's an estimate for a week in the future, assuming that exponential growth continues according to the same average law. Boud (discuss • contribs) 00:52, 19 March 2020 (UTC) (original)
- As I understand it, the idea is that the confirmed cases of, say, day 14 are pretty indicative not of the true cases of day 14 but rather 7 days before, that is, day 7. Therefore, the confirmed cases of day 7 and day 14 would give us an estimate on the true cases of day 0 and day 7. But we want the true cases for day 14. If we assume that the rate of exponential growth between day 0 and day 7 was the same as between day 7 and day 14, we can determine the base of the growth in true cases between day 0 and day 7 (by using confirmed day 7 and day 14 as estimates), and obtain the true cases of day 14 by extrapolation of the exponential growth. Of course, if drastic intervention has curbed the growth somewhere between day 7 and day 14, the assumption would be invalid. And the result of this kind of extrapolation could still easily be underrepresenting the true cases: due to hugely incomplete testing, the confirmed cases of day 14 could be still lower than the true cases of day 7. It seems to be an interesting exercise, to drive home that the true cases may vastly differ from confirmed cases, and to get some kind of first idea. We should keep in mind that incomplete testing impacts not only confirmed cases but also confirmed deaths, although arguably less so. --Dan Polansky (discuss • contribs) 07:36, 20 March 2020 (UTC) (original)
- OK, I see the point. If we count a median of 5 days for incubation, and 1-2 days for a test (depends which country, how well equipped they are, and so on), and assume that all people with suspicious symptoms/travel/contact history are tested, then the lab-confirmed cases on day 14 roughly represent the number of true infections about a week earlier. At least in the PL case, we have the sourced data for the official numbers suspected/hospitalised, quarantined or monitored, starting from 14 days before the first SARS-CoV-2 positive lab confirmation, along with the more common lab-confirmed cases and death counts, but I think there are too many evolving factors to try to infer any serious conclusions from the raw numbers alone. Boud (discuss • contribs) 00:50, 21 March 2020 (UTC) (original)
- Yes, there are many factors to consider, but I wanted to show a more accurate estimate of current infected cases. Some people believe that the number of cases reflect the present, overlooking the time between the infection and the report. --Julian (discuss • contribs) 12:45, 21 March 2020 (UTC) (original)
- OK, I see the point. If we count a median of 5 days for incubation, and 1-2 days for a test (depends which country, how well equipped they are, and so on), and assume that all people with suspicious symptoms/travel/contact history are tested, then the lab-confirmed cases on day 14 roughly represent the number of true infections about a week earlier. At least in the PL case, we have the sourced data for the official numbers suspected/hospitalised, quarantined or monitored, starting from 14 days before the first SARS-CoV-2 positive lab confirmation, along with the more common lab-confirmed cases and death counts, but I think there are too many evolving factors to try to infer any serious conclusions from the raw numbers alone. Boud (discuss • contribs) 00:50, 21 March 2020 (UTC) (original)
- As I understand it, the idea is that the confirmed cases of, say, day 14 are pretty indicative not of the true cases of day 14 but rather 7 days before, that is, day 7. Therefore, the confirmed cases of day 7 and day 14 would give us an estimate on the true cases of day 0 and day 7. But we want the true cases for day 14. If we assume that the rate of exponential growth between day 0 and day 7 was the same as between day 7 and day 14, we can determine the base of the growth in true cases between day 0 and day 7 (by using confirmed day 7 and day 14 as estimates), and obtain the true cases of day 14 by extrapolation of the exponential growth. Of course, if drastic intervention has curbed the growth somewhere between day 7 and day 14, the assumption would be invalid. And the result of this kind of extrapolation could still easily be underrepresenting the true cases: due to hugely incomplete testing, the confirmed cases of day 14 could be still lower than the true cases of day 7. It seems to be an interesting exercise, to drive home that the true cases may vastly differ from confirmed cases, and to get some kind of first idea. We should keep in mind that incomplete testing impacts not only confirmed cases but also confirmed deaths, although arguably less so. --Dan Polansky (discuss • contribs) 07:36, 20 March 2020 (UTC) (original)
Discussion moved to its own page
[edit source]I moved the discussion to this page. I also added the links to the original submissions. --Julian (discuss • contribs) 15:38, 22 March 2020 (UTC)
Base of the confirmed case growth vs. base of true case growth
[edit source]The method used assumes that the base of confirmed case growth is the same as the base of true case growth, which is unlikely to be the case. The base of confirmed case growth critically depends on the base of test count growth. I mean "base" as the base of the exponential growth.
One interesting source of the subject:
- Coronavirus Case Counts Are Meaningless*, fivethirtyeight.com
Pinging Julian.
--Dan Polansky (discuss • contribs) 07:32, 8 April 2020 (UTC)