05-20-2021, 02:28 PM
(05-20-2021, 01:55 PM)megabytephreak Wrote:(05-20-2021, 12:12 PM)jeffster Wrote: I'm going to try to provide actual case numbers here, with adjustments to 1 day prior. As you all know, reporting is very inaccurate, and doesn't tell an accurate story. In some cases, a "confirmed" case being reported could be over a year old. For example, todays numbers included 1,355 cases that occurred May 4-18. 80 cases are from early 2021 and 2020.
Another example, yesterdays positive tests were 965, which might not be fully accurate either when tomorrows result come in, which likely will include some cases from todays numbers.
The episode date (which is the date you are using) is not directly connected to the test date. They must ask when people started feeling sick/last possible exposure. So it's very reasonable that many of the tests have episode dates up to 1-2 weeks prior to the current date. The episode date is more revealing than the test date for epidemiological purposes.
What the data means is that there are 965 cases found yesterday which cannot be tied to an earlier exposure risk/symptom start. Many (but not all) of the remaining cases will have been tested in the last day or two, but had information indicating an earlier case start date. The vast majority of the episode dates seem to fall within the 2 week window for developing symptoms after exposure.
For providing a real-time readout, the number of reported positive tests is still a reasonable metric. You can't have a reasonable readout of the number of transmissions on a date for until somewhere between 1-2 weeks after that date, due to time for symptom development and test processing/reporting. The Waterloo region site allows to view the data by either reporting date or episode date, but notes the lag means the reporting by episode date is under-counted in the immediate past.
https://www.regionofwaterloo.ca/en/healt...egion.aspx
Tomh can confirm this, but I am fairly certain the numbers they are reporting are testing dates. Episode date would result in most cases being added to totals days before. The numbers Tomh reports are mostly day of, but for some reason frequently include backdated corrections.
While episode date has value, over long enough time periods it's equivalent to test date, and more, because symptoms take 2-14 days to appear, it's never particularly accurate in the short term either.
It is worth looking at both, but episode date does not explain the backdated corrections.