Read on Twitter
Thread:
1. Why I have resisted simple test volume adjustments for Ontario's new case curve. It's the fallacy of aggregate assumptions. We know more tests generally means more cases, ALL ELSE BEING EQUAL.
We know Ontario hit a 7 day avg of 12.1% test positivity on April 11
1. Why I have resisted simple test volume adjustments for Ontario's new case curve. It's the fallacy of aggregate assumptions. We know more tests generally means more cases, ALL ELSE BEING EQUAL.
We know Ontario hit a 7 day avg of 12.1% test positivity on April 11
2. April 11th saw 3,569 tests reported and 431 cases.
This was before Ontario changed the way it reports. On April 15th they started counting all tests, including multiple tests on one person. We don't know how to adjust for that. https://www.thestar.com/news/canada/2020/04/16/the-province-has-changed-the-way-it-reports-covid-19-testing-data-heres-why-you-should-care.html
This was before Ontario changed the way it reports. On April 15th they started counting all tests, including multiple tests on one person. We don't know how to adjust for that. https://www.thestar.com/news/canada/2020/04/16/the-province-has-changed-the-way-it-reports-covid-19-testing-data-heres-why-you-should-care.html
3. We know that LTC outbreaks dominated April's reporting. But public reporting only started on April 24.
From press conference updates, we know Ontario geared up to a mass surveillance of the outbreaks, even focusing one week on LTCs. But test numbers were only ~10% of now.
From press conference updates, we know Ontario geared up to a mass surveillance of the outbreaks, even focusing one week on LTCs. But test numbers were only ~10% of now.
4. There are 3 "buckets" we might lump tests into:
- High positivity outbreaks,
- Medium positivity, sporadic pre-screened symptomatics and traced (which Ont did in April), and
- Low positivity mass access, with many asymptomatics (starting in May as capacity improved)
- High positivity outbreaks,
- Medium positivity, sporadic pre-screened symptomatics and traced (which Ont did in April), and
- Low positivity mass access, with many asymptomatics (starting in May as capacity improved)
5. Of course, due to the lack of [public] data, WE DO NOT KNOW THE TEST NUMBERS for these three buckets.
But as a "thought experiment", let's see what happens if we plug in some hypothetical numbers. Your mileage may vary on what you would guess. That's fine.
![5. Of course, due to the lack of [public] data, WE DO NOT KNOW THE TEST NUMBERS for these three buckets. But as a "thought experiment", let's see what happens if we plug in some hypothetical numbers. Your mileage may vary on what you would guess. That's fine.](https://pbs.twimg.com/media/Ei8NbBrWkAITbq2.jpg "5. Of course, due to the lack of [public] data, WE DO NOT KNOW THE TEST NUMBERS for these three buckets. But as a \"thought experiment\", let's see what happens if we plug in some hypothetical numbers. Your mileage may vary on what you would guess. That's fine.")
But as a "thought experiment", let's see what happens if we plug in some hypothetical numbers. Your mileage may vary on what you would guess. That's fine.
6. Where I could, I used the actual aggregate numbers for Sept 22 and April 11 and worked from there to construct this imaginary test summary.
You can see that if the 34,963 actual tests of Sept 22 are applied to April 11th, we come up with 1,590 cases.
You can see that if the 34,963 actual tests of Sept 22 are applied to April 11th, we come up with 1,590 cases.
7. If we just go with a simple adjustment of cases based on test volume growing from 3,569 to 34,963, we get an estimate of 4,231.
We can't use a simple adjustment, it's wildly inaccurate. Would the actual 431 cases on April 11th be 4,231 or 1,590? Or in between? Who knows.
We can't use a simple adjustment, it's wildly inaccurate. Would the actual 431 cases on April 11th be 4,231 or 1,590? Or in between? Who knows.
8. But common sense says it would be closer to 1,590 than 4,231. We need to resist the urge to use the sparse data we are fed and attempt to paper over the gaps with simple adjustments. I would rather compare Euro curves to Canada than go back in time with this little test data.
9. Better yet:
It's time to demand more of Ontario's covid19 test tracking.
It's time to demand more of Ontario's covid19 test tracking.
Why do I spend this much time on what might appear to be an arcane issue?
1) To explain that the April wave peak was higher adjusted for testing but not as high as some "casedemics" distort.
2) To explain why I don't make what look like simple adjustments to my Key Indicators
1) To explain that the April wave peak was higher adjusted for testing but not as high as some "casedemics" distort.
2) To explain why I don't make what look like simple adjustments to my Key Indicators
3) To explain how the lack of proper tracking data on testing hobbles our efforts to track and understand the effects of this virus.
4) To remind folks that there's plenty of current comparative information across the oceans and many are in 2nd waves.
4) To remind folks that there's plenty of current comparative information across the oceans and many are in 2nd waves.
