2020 UK deaths in historical perspective

Most of the above chart is generated from data in ONS' Monthly mortality analysis, England and Wales: Dec 2021 (https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/datasets/monthlymortalityanalysisenglandandwales), Table 10. (Downloaded 21 Jan 2022). In addition I have included data from this 'Freedom of Information' (FoI) report.

Interpretation notes:

The ONS monthly mortality analysis with ready calculated ASMR up to the most recent month for England and Wales only goes back to the year 2001. If we want to examine the longer-term trend we have to calculate the ASMR ourselves. Fortunately, we can get annual death and population figures in standard age groups for the whole of the UK from The Human Mortality Database (HMD, https://mortality.org) covering 1922-2018 (ie since the partition of the island of Ireland). By standardising these to the European Standard Population 2013 we can derive the ASMRs for the whole UK for this period.

It would be strange if the age at which people typically die remained the same but the rate of deaths declined. The fall in mortality rate is, of course, reflected in the rising age at which people typically die.

We can look at age at death another way. What is the life expectancy of a child born in a particular year?

You may wonder whether the observed long-term trend in mortality is unique to the countries of the UK. Wonder no more.

The above chart is rather cluttered; it's difficult to clearly see what's happening to the trend. If we just look at the average of the ASMRs from all nine countries we see the following:

I was looking for a snappy tag-line to describe 2020. Unfortunately, 'The Great Panic' has been used too many times in the past. The best I could come up with was:

2020: The highest annual mortality rate in twelve years.

Although the annual mortality rate for 2020 only reached a 12-year high, most of the extra deaths occurred in a few short periods of time rather than spread throughout the Winter months. It's not too surprising that people found this frightening and worried that the health service would be overwhelmed. Media reports using the term 'exponential' (ie increasing faster and faster without limit) and referring to the 'R Number' or 'doubling every x days' (which are both round-about ways of saying 'exponential') didn't help. Diseases only spread exponentially in Science Fiction; in the real world epidemics follow a Gompertz mathematical function which is essentially two mutually counteracting exponential functions and which is ultimately self-limiting.

The fact that epidemics follow Gompertz functions is somewhat supported by epidemiologists grand modelling theories but was actually just born out of observation. We don't really need to know why, but that they do follow these functions.

We also heard that Sweden (which did not do 'lockdown' like most other countries) fared worse than other Scandanavian countries. For them 2020 was the highest ASMR since... 2013. So, highest mortality in seven years.

The thing we also ought to say is:

2019: The lowest EW annual mortality rate in over 97 years (and probably the lowest ever).

For Sweden 2019 was the lowest ASMR ever.

The peak in 2020 for UK and Sweden was at least partly due to so many frail people not dying in 2019.

Will we learn lessons from this and prepare better when we next have a year with unusually low mortality? Probably not.


An alternative view of age-standardised mortality

One thing that occurred to me as I was researching this post is that it's rather peculiar to measure mortality from 1 Jan to 31 Dec. I know we want a fixed time period for comparisons (one year is usual) but why place the start and end point right in the middle of the usual peak death and public holiday period for the northern hemisphere? Winter months usually bring increased mortality compared with summer months and public holidays can delay death registrations. Also a calendar year is 365 or 366 days. Added to this if we use weekly mortality reports in order to get the most fine-grained information, most years have 52 weeks but some (2009, 2015, 2020) have 53 weeks.

Below I show a chart of mid-year annual (actually 52 week) age-standardised mortality for England and Wales based on the weekly death data published by ONS which (only) goes back to Jan 2010.

Which trend line do you believe?..

...they're all mathematically correct.

I'd like to draw your attention to the fact that because ONS consider 2020 to be an unusual year they have opted to publish 2021 data compared to a 5 year average ending in 2019 (ie they have used 2015-2019). Using 2015-2019 for our comparisons gave the most crazy extreme results of all in our chart above (note they're using an average, I'm using a trend). If the 2015-2019 trend continued we would have no deaths at all by December 2090. Using the 2014-2018 trend gave us something nearly as bad but in the opposite direction.

This note is quoted from the ONS weekly mortality report:

This average is based on the actual number of death registrations recorded for each corresponding week in 2015 to 2019. Moveable public holidays, when register offices are closed, affect the number of registrations made in the published weeks and in the corresponding weeks in previous years. The five-year average has been provided for 2015 to 2019 (rather than 2016 to 2020) because of the impact of the coronavirus (COVID-19) pandemic on deaths registered in 2020. The average for 2015 to 2019 provides a comparison of the number of deaths expected per week in a usual (non-pandemic) year. Data by week for 2020 has been provided separately for comparison.

If you wish to comment you may email me at: SoundOfReason0 at gmail dot com. If I append your comment I will not publish your email address.