Naked Science Forum
Life Sciences => Physiology & Medicine => COVID-19 => Topic started by: katieHaylor on 11/02/2021 11:02:03
-
Paul says:
Please can you explain to me how the Covid-19 statistics are done? In East Lindsey I think we have about 148k people. Therefore to generate a figure per 100k people it must be very extrapolated. How can you ever be sure a test group is representing a major group?
I see Lincoln has dropped massively whilst we went up then down a small %. I am no scientist but as an ex-engineer I can’t help thinking the stats are just a best effort guess. Lots I know have never had a test. Extrapolation cannot be accurate.
-
Always read the small print! "Confirmed cases" per 100k means exactly what it says, and is almost certainly an underestimate. You might get closer to the current actual value by extrapolating the most recent trend, but in the immortal words of Louis B Mayer "Prediction is difficult, especially about the future."
The only absolute fact you can glean from newspaper and TV reports is "the death toll has risen", but you knew that anyway.
-
In East Lindsey I think we have about 148k people. Therefore to generate a figure per 100k people it must be very extrapolated.
To generate a figure for East Lindsey per 100k people, divide the number of diagnosed cases (eg with a PCR test) and deaths by 1.48.
- That is not extrapolation, that is normalisation (to a base figure of cases per 100k population)
- Normalisation is much more accurate than extrapolation
I see Lincoln has dropped massively whilst we went up then down a small %.
When you are dealing with a small number of cases (eg 1 suburb, or a small town), statistical fluctuations cause the figures to go up and down quite randomly.
- This is less of a problem with a large city or a large geographical region, where the case numbers are larger
- For small populations, you have to average over a longer period of time to get the number of cases up to a reasonable number
- This period of time should be at least 1 week, since testing and reporting processes tend to take a day off on the weekend
I can’t help thinking the stats are just a best effort guess.
It is true that most cases of SARS-COV2 infection have no symptoms, and these people are unlikely to be tested.
- However, people who are experiencing respiratory symptoms are likely to be tested
- And people entering a hospital with respiratory symptoms will definitely be tested
So the known and diagnosed cases is a reasonable proxy for the (larger number of) undiagnosed and unknown cases in the community.
How can you ever be sure a test group is representing a major group?
By doing random screening for antibodies, it is possible to extrapolate from the known number of symptomatic & confirmed cases to the unknown total number of infections
- The total number of infections is important to see if/when we might be approaching some level of herd immunity
- But the daily news just reports the known & confirmed cases
Spoiler alert: many more people will die before we achieve herd immunity through natural infection (so get vaccinated as soon as you are eligible!)