Excess mortality 2018 – 2021

England & Wales monthly excess mortality by quinary age band & sex

There are several different ways to go about calculating excess mortality, the most common being to derive a 5-year baseline based on mean values for the preceding period and subtract this from the current period. This is easy to accomplish but has its disadvantages in that seasonal illness doesn’t always erupt in the same month each year and even a span of 5-years is no guarantee of being ‘representative’. In fact, when we think about it, the issue of a representative span of years in something as dynamic as disease and death is a bit like asking which angle of the swing of a pendulum in a grandfather clock is representative of pendulum motion.

Let us put all that griping aside and opt for the common approach of excess mortality as derived from 5-year mean baselines. In the process of doing so we find a rather large fly buzzing in the ointment and that is excess mortality for the year 2021 will end up being derived using the base period of 2016 - 2020. The Office for National Statistics (ONS) get round this by adopting 2015 – 2019 as the base period for 2021 as well as 2020 and I shall do the same.

Just to be clear what is going on here, we’ll have a little recap on methodology…

I’ve derived monthly counts of deaths by quinary age band and sex by date of death using ONS datafiles generated by FOI requests. I’ve then derived monthly population estimates by quinary age band and sex by stitching together datafiles held by ONS, and interpolating and modelling values where necessary. Mortality (deaths per 100k population) has been derived using these two sets of figures. We should note that mortality rates are not per 100k of the entire population but per 100k of the sample population; that is to say, the mortality rate for 50 – 54y male deaths in Jan 1970 has been derived using the count for 50 – 54y male deaths for Jan 1970 divided by the population estimate for 50 -54y male deaths for Jan 1970. Hence the figures I’m throwing about account for population growth by age and sex. Let’s start with a combi slide for the younger males…

The key features here are the pre-pandemic period where excess mortality bobbed about zero, the pandemic period where spikes are observed for 40 – 44y males only during the first and third waves and the peculiarly rapid descent into negative excess for males over the age of 14y. As we have seen this is almost certainly due to failure of the ONS to process death certificates for this group. Using our eyeballs alone we find that the pandemic didn’t figure for males under 40 years of age.

The impact of the pandemic for older males becomes blindingly obvious, though there are two additional things to note here. Firstly, the dip for Dec 2021 may be due to delays in death certificate posting and secondly, we have no idea of what excess mortality looks like for years prior to 2018. Is that spike of April 2020 totally unique or have we seen spikes like this in the past? In future newsletters I shall unveil historic trends dating back to 1975 to provide some perspective.

The first thing to note about this slide for younger females is that the y-axis scaling has dropped from ±8 deaths per 100k for young males to ±5 deaths per 100k, thus we see less variation in excess mortality for younger females. These scales are driven by the pandemic peak and I must say I find it somewhat puzzling that a novel virus should seek out to clobber younger men in preference to younger women. There is a notable dip in mortality for 0 – 4y females from Jul 2010 through to Mar 2021 that finds a faint echo in 0 -4y males and it is worth cogitating why. Again we find evidence of delays on death certificate processing, particularly for 40 – 44y females.

It’s hard to distinguish between this slide for older females and the slide for older males but excess mortality rates for men are slightly higher. Worth mentioning are the negative excess dips for the winter season of 2018/19 and 2019/20 which indicate milder flu seasons than usual, which invariably leads to the ‘tinderbox effect’. The dip for Mar 2021 is also intriguing, and worthy of further munching.