9 Comments
User's avatar
Freecumbria's avatar

Thanks for these analyses Clare, very interesting.

You can fill in the covid deaths in the younger age groups by subtracting the non-covid deaths from the all cause deaths in table 2. Don't use table 4 as there are more unknowns because of the male/female split. So if all cause deaths for an age group and month are 14 and non-covid deaths are 13, you can replace the <3 covid deaths with 1.

This leaves some <3 figures left but you can make guesses of these by looking at neighbouring months and comparing person years and deaths in these. Many are likely to be zero. The guesses aren't that material as long as your estimate is remotely sensible.

Should be possible to do 18-39 and 40-49 age groups as a result.

Expand full comment
Clare Craig's avatar

Excellent -thank you!

Expand full comment
Freecumbria's avatar

My 40-49 covid death estimates are here

https://ibb.co/6tPJxrG

Expand full comment
Freecumbria's avatar

The 18-39 covid death figures I've estimated are here

https://ibb.co/g3rx6RX

Expand full comment
Clare Craig's avatar

I did try this. I found that there was still enough ambiguity in the young because of the <3 figures that the covid mortality wasn't reliable. We need better data.

Expand full comment
Freecumbria's avatar

I've not really analysed the difference between table 2 and table 5 covid mortality so I sensed the estimations might be good enough but I can see why they might not be. An example of the difficulty is that the table 2 18-39 June 2021 ever vaccinated covid deaths total 6 deaths (122 - 118 + 28 - 26), before you count the <3 ones, but that's more than the table 5 total of 5!

More generally I also use the 10+ totals (in tables 1 and 3) to make sure I've not over-estimated the missing <3s in over the 18+ groups through the 18+ age group summing to more than the 10+ total, I remember I had to take out a few assumed covid deaths out from the 18+ groups when I did that. You can also calculate the 10-17 all cause mortality and it gives a sensible result but that's super sensitive to filling in the <3 figures.

On the subject of better data, have you noticed there is an error in the ONS female June and July 2021 all cause 10+ ASM figures in table 3, and in the corresponding person 10+ ASMs in table 1. This chart shows it's almost certainly an error.

https://ibb.co/1KrWrGN

Expand full comment
harry's avatar

Thanks for these interesting analyses as well!

If I'm not mistaken then there's a little error in your summary introduction.

Starting with a little nitpicking: it looks to me that you show not the average but the totals of each group over 2022 (basing myself on your monthly graphs for 70-79 in your earlier article). Correct?

In the 80+ all cause mortality graph for 2022, there the yearly totals don't seem to match your monthly rates, for example vaxed ONS av.800x12= ca.10'000/yr whereas your graph shows 20'000/yr.

Expand full comment
Clare Craig's avatar

Thank you!

Yes - I had got the denominator wrong and have corrected. Thank you.

Expand full comment
Graphite's avatar

Thanks for analysing this Clare...

Expand full comment