THE BBC wrote large: ‘The Covid pandemic . . . excess deaths . . . highest level . . . since World War Two . . . 85,000 more than would be expected . . . an increase of 14 per cent . . .’.
Well, is it true that 85,000 have died early because of Covid? To get a better idea we can look again at the graph I left you with last time, which showed how the Age Standardised Mortality has varied recently. Thanks to links kindly provided by those who have commented, this graph is produced directly from ONS statistics and goes back to 2001. On this occasion I have averaged the results over a three-month period.
If I hadn’t lived through 2020, I don’t think the figures would point me to it as the obvious year which contained a pandemic.
For a start, from 2001 to 2011 there is a steady improvement in mortality, levelling out after 2011 (things I’m going to leave you to comment on). The figures show 2020 to be significantly below some relatively recent years.
The annual rise and fall of deaths
2020 has a kink on the way up but it shares with other years the repeated winter rises and summer falls in mortality. We don’t entirely know why these happen. It may be to do with the transmission of respiratory diseases. Perhaps ultraviolet from the sun destroys viruses in aerosols. Or it may be that the low sun means that many of us are so short of vitamin D that it leads to the increased mortality.
The oscillating nature of winter deaths
While summer death rates are fairly consistent, winter death rates oscillate widely. Many of you remarked on the way that particularly high, or low, winter deaths are followed by the reverse a year later. One can see reasons for this. Most deaths, at any time, are those who are vulnerable. If the death rate in one year is low, fewer of the vulnerable will die in that year, leaving more vulnerable to die in the following year, and therefore a higher death rate. In a year when, say, the flu vaccine is poorly matched to the prevalent strain of flu, the death rate will be high: fewer of the vulnerable will survive to the following year and the death rate then will fall.
Here are Age Standardised Mortality rates from 2001-2020:
You can see the oscillating death rates – the 2009 swing from 0 per cent down to minus 5.3 per cent, the 2016 swing from 4.2 per cent to minus 2.6 per cent, etc. What is pretty clear is that previous years’ rises and falls are bad predictors of the following year because of the oscillating nature of winter deaths.
Fair comparisons don’t use raw figures
This table shows that the mortality for 2020 was 12.8 per cent more than 2019 – less than the 14 per cent figure put about by the BBC. Is the BBC figure wrong? Well, no. But the BBC is comparing raw death figures and such a comparison ignores the fact that the population in 2020 is both larger and older than in 2019. The only way to compare one year fairly with another is to use standardised mortalities.
But what about the figures that are trotted out every week (‘compared with the averages of the previous five years’) – are they not raw figures?
Yes, indeed. Every time comparison figures are given to us by the media or the government they use raw comparison figures, not Age Standardised Figures. This makes increases in death rate seem bigger than they actually are. As for the rise from 2019 to 2020, it is significant. But before deciding how significant, we need first to consider what death rate for 2020 we should have expected.
Last year’s death rate should not be used to predict this year’s
Here are Age Standardised Mortality rates plotted for each year. The solid red line shows the actual figures. The dotted blue line shows the predictions based on ‘the average of the previous five years’. You can see how poor are the predictions – the 2019 prediction was about 30,000 deaths more than actually occurred.
How might we make a good prediction?
It’s homework time again. You may have noticed that in Figure 3 I included the government-predicted figure for 2020 but left off the actual figure. That’s to give you a chance to look at the red line and make your own prediction. Imagine it had been your job to produce an expected figure for 2020. How would you have done it? What figure would you have given for the 2020 mortality and does it result in you agreeing with the BBC that there were 85,000 excess deaths in 2020? You might then like to come up with an expected figure for 2021 deaths.
I’ve switched to 4000 iu vitamin D, so this week’s prize is my old bottle of 1000 iu tablets if you get closer than me.
