Coronavirus - Data Analysis Thread
Discussion
Mini update. This compares todays data with the same period last year.
In 2020 we were already seeing cases per 100k tests surge due to the Alpha wave in London and Kent. So far in 2021 everything is holding pretty steady with admissions and deaths at much lower levels.
If Omicron is going to have a serious impact, it should be visible here in cases very soon
In 2020 we were already seeing cases per 100k tests surge due to the Alpha wave in London and Kent. So far in 2021 everything is holding pretty steady with admissions and deaths at much lower levels.
If Omicron is going to have a serious impact, it should be visible here in cases very soon
I have been playing around with the numbers a bit based on the 'probable' omicron cases detected through S gene drop out on PCR surveillance and reflecting UKHSA's view that we may have had 200k Omicron infections on 13th Dec.
You can get to current levels of cases with some basic assumptions:
I think there is little we can do now other than watch the daily cases. We ought to be at around 100k today, but reporting lag means we won't know till next week.
On these numbers we hit 26m Omicron infections by Christmas day, so if this is in any way representative of what we are about to go through, its not going to be sustainable for long.
You can get to current levels of cases with some basic assumptions:
I think there is little we can do now other than watch the daily cases. We ought to be at around 100k today, but reporting lag means we won't know till next week.
On these numbers we hit 26m Omicron infections by Christmas day, so if this is in any way representative of what we are about to go through, its not going to be sustainable for long.
It's update day again.
1. Comparison to last year. This is showing Cases per 100k tests staying flat, whereas this time last year they were increasing rapidly. This is perhaps counter intuitive given the reported numbers, but testing has really surged:
2. Tests and Cases. Its fascinating to see that the recent increase in raw cases is exactly in line with the recent increase in testing, which is now hitting 1.6m per day. Test positivity, which I believe to be the best indicator of prevalence, has not changed.
3. There is no substantial change in terms of Cases per 100k Tests or Admissions. The corresponding Case Fatality Rate and Case Admissions Rate remain low.
4. This graph shows Cases per 100k test, Admissions and Deaths. The close relationship between these metrics is why I believe it is appropriate to correct cases for testing volumes and focus on positivity rates.
1. Comparison to last year. This is showing Cases per 100k tests staying flat, whereas this time last year they were increasing rapidly. This is perhaps counter intuitive given the reported numbers, but testing has really surged:
2. Tests and Cases. Its fascinating to see that the recent increase in raw cases is exactly in line with the recent increase in testing, which is now hitting 1.6m per day. Test positivity, which I believe to be the best indicator of prevalence, has not changed.
3. There is no substantial change in terms of Cases per 100k Tests or Admissions. The corresponding Case Fatality Rate and Case Admissions Rate remain low.
4. This graph shows Cases per 100k test, Admissions and Deaths. The close relationship between these metrics is why I believe it is appropriate to correct cases for testing volumes and focus on positivity rates.
Mahalo said:
If you increase testing by large amounts as the UK has done in the last three weeks and test positivity remains the same then that means that prevalence has increased.
We only ever see part of the epidemic, depending on our testing infrastructure. The ONS infection survey consistently shows more infections than the cases we identify. UKHSA's recent analysis of omicron suggested that cases were aound 40% of infections. So there is plenty of opportunity to see more of what is going on by increasing testing levels. Our testing is a mix of symptomatic and asymptomatic people. If you increase the number of people tested, but keep that mix the same, then you would expect cases to increase proportionately.
That is exactly what has just happened.
If prevalence was increasing, then a higher proportion of the sample (whatever the sample size) should test positive. We have seen this in recent waves. South Africa has seen it with omicron. We are not seeing it right now, but I suspect that will change.
RSTurboPaul said:
Elysium said:
It's update day again.
1. Comparison to last year. This is showing Cases per 100k tests staying flat, whereas this time last year they were increasing rapidly. This is perhaps counter intuitive given the reported numbers, but testing has really surged:
<snip>
An idle thought, but do the flat rates of deaths and admissions now reflect endemic status?1. Comparison to last year. This is showing Cases per 100k tests staying flat, whereas this time last year they were increasing rapidly. This is perhaps counter intuitive given the reported numbers, but testing has really surged:
<snip>
The fact it has barely changed despite seasons changing and population behaviour (schools, work, etc.) and 'variants', and the fact it has been the same despite crossing over the 2020 lines at the end of September, seems to suggest to me that this is potentially going to be the ongoing rates that we'd expect as 'background' Covid?
I'm not sure how that would fit with the ultra-low figures last summer, though, other than perhaps there really was very little about at the time.
That is exactly what SARS-CoV-2 has been doing in the UK over the last 4 months.
the-photographer said:
Before we run out of testing
Still around 2.1-2.2 day doubling in the latest Omicron report. Now up to 73.5% SGTF in London. (Which may actually be good in a weird way because it means that a fair proportion of the future pain is already priced into the case ratios we are now seeing).
https://twitter.com/BristOliver/status/14715303987...
This is the rate at which Omicron might be replacing Delta. Still around 2.1-2.2 day doubling in the latest Omicron report. Now up to 73.5% SGTF in London. (Which may actually be good in a weird way because it means that a fair proportion of the future pain is already priced into the case ratios we are now seeing).
https://twitter.com/BristOliver/status/14715303987...
What's not clear to me at this point is the extent to which this means additional cases.
I can see how it can very quickly outcompete a less transmissible variant, but I am doubtful that it can grow at the same rate in the wider population.
fat80b said:
Elysium said:
Just wondering what I'm doing wrong ?
I have newCasesByPublishDate, newCasesBySpecimenDate, and newPCRTestsByPublishDate which I think should allow me to plot almost the same graph but the numbers are completely different - e.g. This gives ~10K per 100K for all days last week where your graph has 4/100 ?
There is a download link at the bottom left of the first graph here:
https://coronavirus.data.gov.uk/details/testing
To calculate Cases per 100k tests I use a 2 day lag between Cases by Specimen Date and Tests by Published Date. This allows for the typical lag between giving a specimen and having it processed.
Current tests are around 1.6m and cases around 90k. So the calculation is 90,000/(1,600,000/100,000) = 5,625
I had a play around with some numbers last week to try to make sense of the 200k infections figure Javid stated in Parliament on the 13th Dec:
https://news.sky.com/story/covid-19-why-health-sec...
This was my projection:
This is how it looks compared to actual cases by specimen date, noting that these are still subject to reporting lag:
The rise in cases does support the idea that omicron is surging. However, it is very clear that it is also driven by a dramatic increase in testing volumes.
If the SAGE exponential growth argument is in any way valid, we should be expecting something in the order of 500k cases to be reported today.
https://news.sky.com/story/covid-19-why-health-sec...
This was my projection:
This is how it looks compared to actual cases by specimen date, noting that these are still subject to reporting lag:
The rise in cases does support the idea that omicron is surging. However, it is very clear that it is also driven by a dramatic increase in testing volumes.
If the SAGE exponential growth argument is in any way valid, we should be expecting something in the order of 500k cases to be reported today.
Nice South Africa data dashboard here:
https://mediahack.co.za/datastories/coronavirus/da...
Cases seem to be peaking in all regions now and Rt is falling:
https://mediahack.co.za/datastories/coronavirus/da...
Cases seem to be peaking in all regions now and Rt is falling:
Todays reported cases are below 100k which conclusively disproves the SAGE 'doubling every 1.9 days' model
1. Testing has dropped back a little and is now around 1.4m per day. As a consequence, we can now see that cases have accelerated a bit faster than testing. So we have general increase in positivity:
2. You can see this in the comparison with last year
3. And perhaps more clearly here
1. Testing has dropped back a little and is now around 1.4m per day. As a consequence, we can now see that cases have accelerated a bit faster than testing. So we have general increase in positivity:
2. You can see this in the comparison with last year
3. And perhaps more clearly here
Update
1. Comparison with last year, which I have now extended to the end of January. There is a definite rise in cases per 100k tests. However the rate of change is very similar to the Alpha wave we saw as we exited the circuit breaker last year. Admissions still about half of the level we saw last year and deaths almost 5 times lower:
2. You can see the impact of rising testing here. However, the gap between these two lines is narrowing indicating that cases are rising faster than tests. We are seeing a 30% increase in the 7 day rolling total of cases per 100k tests. Nowhere near the 1.9 day doubling initially suggested by SAGE.
3. You can see the growth in Cases per 100k tests in context here. We are still below the 'euro final peak' that occured just before freedom day. The Case Fatality Rate is falling as these new cases have not yet resulted in more deaths (0.23% over the last 14 days). The Case Hospitalisation Rate is also gently falling (1.77% over the last 14 days)
4. The first graph compares these three metrics since 1st March with a lag for admissions and deaths and deaths rebased to align the curves visually. Admissions look as if they might be increasing in response to case growth. Deaths so far are not.
The second graph shows deaths without any rebasing and includes last winters wave for context. This continues to show that the three metrics remain aligned, but that cases now result in proportionally fewer admissions and deaths.
1. Comparison with last year, which I have now extended to the end of January. There is a definite rise in cases per 100k tests. However the rate of change is very similar to the Alpha wave we saw as we exited the circuit breaker last year. Admissions still about half of the level we saw last year and deaths almost 5 times lower:
2. You can see the impact of rising testing here. However, the gap between these two lines is narrowing indicating that cases are rising faster than tests. We are seeing a 30% increase in the 7 day rolling total of cases per 100k tests. Nowhere near the 1.9 day doubling initially suggested by SAGE.
3. You can see the growth in Cases per 100k tests in context here. We are still below the 'euro final peak' that occured just before freedom day. The Case Fatality Rate is falling as these new cases have not yet resulted in more deaths (0.23% over the last 14 days). The Case Hospitalisation Rate is also gently falling (1.77% over the last 14 days)
4. The first graph compares these three metrics since 1st March with a lag for admissions and deaths and deaths rebased to align the curves visually. Admissions look as if they might be increasing in response to case growth. Deaths so far are not.
The second graph shows deaths without any rebasing and includes last winters wave for context. This continues to show that the three metrics remain aligned, but that cases now result in proportionally fewer admissions and deaths.
Ashfordian said:
Elysium said:
I don't think this level is localised enough. Some areas have peaked while the outer boroughs are still seeing rising cases. What doubling rates do you get if you drill down to Lambeth and Wandsworth LA level?It looks like the spike in these LA's has passed for now and the rate of cases per 100k is about double or a smidge over double, compared to what we achieved with previous variants (~3.1k verses ~1.5k).
Those explosions overlap to create a bigger wave.
This is Lambeth. It briefly hit a 3 day doubling, but has very clearly already slowed:
Wandsworth looks very similar.
Ashfordian said:
Elysium said:
COVID grows explosively fast at local level, but slows very quickly. Like a series of fireworks.
Those explosions overlap to create a bigger wave.
This is Lambeth. It briefly hit a 3 day doubling, but has very clearly already slowed:
Wandsworth looks very similar.
I agree with you about a series of fireworks if you look too local. These are 330k population areas though which should be large enough to give a worst case scenario.Those explosions overlap to create a bigger wave.
This is Lambeth. It briefly hit a 3 day doubling, but has very clearly already slowed:
Wandsworth looks very similar.
I know this is not a standard scientific method, but it looks like Omicron is about 2-2.5x more transmissible than previous variants based on the 7 day case/100k rates
The data is going to be really inconsistent over the next 3 weeks so I'm not sure we'll have any clarity now until mid-Jan
NWTony said:
Ashfordian said:
the-photographer said:
I wonder if the same agency will publish what the reality compared to this graph.
Or will they simply try to sweep their incompetence under the carpet? Again!
https://www.ons.gov.uk/peoplepopulationandcommunit...
Update
1. Comparison with last year. The growth rate in cases per 100k tests is almost exactly the same as last year and the current level is now very similar to 23rd Dec 2020.
Although admissions are noticeably increasing they are holding at about half the level we saw then. Deaths are holding steady and are now about 9 times lower than Dec 23rd 2020:
Due to the complexities around 'with' and 'due to' COVID admissions, I have added a new graph looking at numbers in hospital and MV beds occupied. The number of cases in hospital and MV beds remains about half of the level we saw last year and most importantly, is not increasing at the same rate as admissions:
2. This shows tests and cases. We are now running at 10 million tests per week. 1.5 million a day or about three times more than this time last year. However, you can see that cases have risen faster than tests, which indicates a genuine increase in prevalence:
3. These graphs show Cases per 100k Tests with the associated Case Fatality Rate (CFR) and Admissions with the associated Case Hospitalisation Rate (CHR). Although cases and admissions are increasing the proportion of cases going to hospital or dying is falling. The CFR is averaging 0.2% over the last 14 days and the CHR has fallen to 1.6% over the same period:
4. The first graph shows Cases per 100k Tests, Admissions and Deaths overlaid, with a lag and rebasing of death numbers to align the curves. We can see that admissions is now rising in line with Cases per 100k Tests, but so far deaths are not following this trend. The second graph shows these metrics over a longer period, with raw numbers for deaths and shows how different our current situation is to the one we faced last year:
1. Comparison with last year. The growth rate in cases per 100k tests is almost exactly the same as last year and the current level is now very similar to 23rd Dec 2020.
Although admissions are noticeably increasing they are holding at about half the level we saw then. Deaths are holding steady and are now about 9 times lower than Dec 23rd 2020:
Due to the complexities around 'with' and 'due to' COVID admissions, I have added a new graph looking at numbers in hospital and MV beds occupied. The number of cases in hospital and MV beds remains about half of the level we saw last year and most importantly, is not increasing at the same rate as admissions:
2. This shows tests and cases. We are now running at 10 million tests per week. 1.5 million a day or about three times more than this time last year. However, you can see that cases have risen faster than tests, which indicates a genuine increase in prevalence:
3. These graphs show Cases per 100k Tests with the associated Case Fatality Rate (CFR) and Admissions with the associated Case Hospitalisation Rate (CHR). Although cases and admissions are increasing the proportion of cases going to hospital or dying is falling. The CFR is averaging 0.2% over the last 14 days and the CHR has fallen to 1.6% over the same period:
4. The first graph shows Cases per 100k Tests, Admissions and Deaths overlaid, with a lag and rebasing of death numbers to align the curves. We can see that admissions is now rising in line with Cases per 100k Tests, but so far deaths are not following this trend. The second graph shows these metrics over a longer period, with raw numbers for deaths and shows how different our current situation is to the one we faced last year:
Edited by Elysium on Friday 31st December 17:27
the-photographer said:
Might be hard spotting the peak using cases, have to use hospital admissions 5/7/14 days later
https://twitter.com/russss/status/1476258248042942...
That’s not up to date. https://twitter.com/russss/status/1476258248042942...
There is a ‘real time’ version here:
https://russss.github.io/covidtracker/testing.html
Gassing Station | News, Politics & Economics | Top of Page | What's New | My Stuff