Dr No recently came across a number of reports in the mainstream media suggesting that the median age of those dying from covid–19 was a little higher than the median age for all deaths: 82.4 for covid deaths, against 81.5 for all deaths. Putting aside the inevitable journomonologophobia that gave rise to a swirling word salad of averages, means and medians in the reports, the finding gives rise to the clinically intriguing possibility that the best way for elderly people to live longer in 2020 was not to bother with all that shielding nonsense, but instead get out there and catch the damn thing. On more sober reflection, Dr No wonders if one or more biases may be at work, but let’s for a moment consider that the reported findings are true. If they are, it gives yet more weight to the notion that the burden of covid–19 falls with greatest weight on the very old. If the median (middle value when all are ranged from top to bottom) age for covid deaths is 82.4, then half of all covid deaths occur in those aged 82.4 and over.  

All of the reports, including those in the Times (paywall), the Daily Mail and the Sun, give their source as a study done by the Centre for Evidence Based Medicine in Oxford. This seems plausible, given the CEBM’s penchant for pricking  pandemic balloons, a plausibility reinforced by the detailed further and better particulars in the reports on the numbers from the study. Hot on the trail, Dr No set off into Googleland with all his favourite search tricks to hand — and drew a complete blank. Unless a reader manages to find the original study, it appears it has never been published. It is possible, Dr No supposes, that the CEBM put out a press release and nothing else, but that would be highly unusual. So, once again, it looks as though we have yet another bummer number. Given that we have no way of verifying the data and methods used, we have no way of knowing whether the figures reported, 82.4 and 81.5 are fact or fantasy.

So what are the real median ages for covid deaths and all deaths? After more time in Googleland, Dr No came across this response from ONS to a request for the covid element of the question. Based on deaths up to 2nd October, the median age for all covid deaths in England and Wales was 83 (81 for males, 85 for females). The means were also reported, at  80.4, 78.7 and 82.5 respectively, but not the mode (commonest age at death). It seems that, even though the source appears ephemeral, the 82.4 figure is near enough, in the horseshoes and hand grenades sort of way, but we’ll use the ONS figure as we have a direct source, and say 83 is the median age for covid deaths.

Median age at death is only one measure of longevity, and is not routinely reported by ONS. Instead, it tends to prefer life expectancy, a vaguely mind bending sort of mean statistic based on a little light touch numerology. What exactly do we mean if we say, for instance, that life expectancy today is 80 years? Another measure is the mode, or commonest value. All of these are stabs at summarising in a single figure where the typical, in the middle value lies. Means (averages) work well for Normally distributed (the symmetrical bell shaped curve) variables, but fail, often spectacularly, with skewed (asymmetrical) distributions, including age at death, which is heavily skewed to the right (most people die when they are old). For these heavily skewed variables, the median (middle value when all are ranged top to bottom) gives a better sense of the typical, in the middle (which it literally is) value.

ONS do publish annual tables of deaths by single year of age for England and Wales, and these are more than enough to provide a very good estimate of the median age at death. Because these tables are aggregates of all deaths, all we need to do is find the year that includes the middle (total deaths/2) death. What we find from the latest data (for 2018) is that for all persons, the median age at death is 80-81 (nearer 81, lets say 81.7). As the tables also  report deaths by sex, we can do the same for males and females, and note that for males the median at death is  78-79 (somewhere in the middle, lets say 78.5) and for females it is 83-84 (somewhere in the middle, lets say 83.5).

It turns out that, even if the CEBM research is ephemeral, the findings are not far off the verifiable figures, which are if anything even more striking. The alleged CEBM median ages are 82.4 (covid deaths) and 81.5 (all deaths), whereas Dr No’s verifiable figures based on ONS data 83 (covid deaths) and 81.7 (all deaths). Both medians are marginally higher, but more striking is the difference between the covid and all cause medians, which has risen from 0.9 years to 1.3 years. Folks with covid tend to live over a year longer than those without. Perhaps covid really is good for you, after all!

Of course it is not. There is almost certainly a selection bias at work here. What we might expect, without the bias, is that covid pushes frail birds off their perches a little early, so they die a little early. But this does not appear to happen. One possible, indeed likely, explanation is that covid selects (the bias) the really really frail, and to be really really frail you have — by and large, we are talking at population level here  — to be really really old. The median age at death for covid is higher because, by and large, you need to be very old, and very frail, for covid to stand a reasonable chance of pushing you off your perch.

The two by and large qualifications in the last paragraph are there because of course there are exceptions, just as there is long covid, but by and large the brutal fact is that fatal covid is overwhelming a disease of the very old. This is borne out by other observations. The ONS weekly data consistently show — this is the easy to remember one — that 75% of covid deaths occur in those aged 75 and over, and that getting on for half (42%) of the deaths occur in those aged 85 and over, a figure fully consistent with a median age at death of 83 years old. The same data, by the way, show that for children up to and including the age of fourteen — so very much the schoolchildren so often in the news these days as biological covid bombs — the risk of fatal covid is almost vanishingly small, with only seven deaths, very likely in those with pre-existing conditions, occurring in this age group.    

Though grim for most of the very old, though not all, because some will feel themselves ready to take the final journey, these findings are in another way a cause for optimism. To the extent that fatal covid is very much a disease of the old, then it is also not a disease of the young and middle aged. And buried within all this, though hard to prove at this stage, as we lack complete data, there is another possible cause for optimism, and indeed explanation of why the current seasonal increase in deaths is not as bad as the spring peak — the latest seven day moving average peak, in late November, topped at just under 500 deaths, compared to just under 1,000 in the spring. If fatal covid overwhelmingly attacks the very old and frail, then on its first arrival it will operate very much in the same way that a first, or prevalence, screening round works: there will be many cases/fatalities, because of the larger pool, accumulated over time, of pre-clinical/susceptible people. Subsequent screening rounds, or infection waves, affect fewer people, because the pool is now only made up of new (incident) pre-clinical/susceptible people that have appeared since the last screening round or infection wave. There is simply, to use the incendiary phrase, less dry tinder in the grate, and any subsequent fires will be smaller.

The hard facts, and these are hard facts, not scenarios, projections, estimations, models, lottery numbers, astrological predictions or numerological fancies, of the age distribution and median age for covid deaths are brutal facts, particularly for the very old and their family and friends. But they also provide a much needed sense of perspective, of what covid really is, and, just as importantly, what it is not. We need that perspective now, more than ever, if we are to move back to the old normal, and a practical, sensible and above all else proportionate response to this novel seasonal flu like illness. 


  1. Shawn Gibson Reply

    I’m fond of saying that we’ve forgotten more about Covid than we’ve learnt. We used to know that it was a disease of the very elderly. Everyone knew that. When it existed only on the news in China we all knew this and did not worry. I remember Clare Gerada tweeting that it was a disease of the elderly and we needed to learn to live with it – I don’t think she would dare tweet that now. There’s much else we’ve forgotten of course, like physical distancing and basic hygiene helps but little else does. Like herd immunity is a thing and was a thing long before Covid. The general public have no hope of knowing what to believe any more.

  2. dearieme Reply

    The fact that it’s the ancient who get first dibs on the vaccine tells the same story. Also I suppose they are not too much at risk of long term side effects.

    Though thinking further, isn’t it particularly old, male fatsos who are at greatest risk? Shouldn’t they get priority?

  3. dr-no Reply

    Shawn – have taken the liberty of adding your correction to your comment, hope that is OK. Yes, CG and other will find it very hard to speak their mind these days – assuming that is they have a mind. Dr No doesn’t mean that in a derogatory sense, more that in the face of a relentless barrage of (mis-)information and/or selective reporting over several months, even the sharpest of minds can become blunted. It takes active ongoing and equally relentless effort to resist the constant narrative. And then there’s the conformity thing: sticking your head above the battlements risks the Milk Curdler turning her blow lamp on you, with deleterious consequences.

  4. dr-no Reply

    djc – thanks for pointing them out. Have just run the same calculations and the results are the same years of age, maybe for all persons has increased to 80.8.

    PS have taken the liberty of editing the link to stop the overflow

  5. Tom Welsh Reply

    “It turns out that, even if the CEBM research is ephemeral, the findings are not far off the verifiable figures, which are if anything even more striking”.

    Is it possible that the CEBM simply did the same as you? Or did they cite a study?

    While this article is very instructive and interesting, I feel we ought to bear in mind that anything involving “Covid-19 deaths” is built on sand. Apart from one intrepid doctor in Germany who did a dozen or so post-mortems, such procedures seem to have been neglected when they weren’t actually forbidden 9as I think they were in the UK). So presumably the only way of identifying a “Covid-19 death” is the thoroughly discredited “had a positive test” criterion. But the tests are unreliable to say the very least…

  6. dr-no Reply

    Tom – no study cited, nothing detectable via google despite thorough searching, or on the CEBM website, but very likely they used the same methods, or something very similar, on a similar but not the same dataset.

    Agree about the over-attribution of deaths to covid, have mentioned this in posts passim. The ONS definition of a covid death is very broad*, and as you say most studies that have looked at the question of co-morbidities and/or real cause of death (not sure PMs are verboten here, perhaps just avoided) have usually suggested the percentage of true covid deaths among all so-called covid deaths is in single figures. In the absence of data, we can only speculate on what the median age at death among true covid deaths might be, so probably best to stick with the data and conclusions we have…

    * The ONS definition of covid deaths gets blurry from time to time. The current weekly report definition is “Coronavirus (COVID-19) deaths are those deaths registered in England and Wales in the stated week where COVID-19 was mentioned on the death certificate. A doctor can certify the involvement of COVID-19 based on symptoms and clinical findings – a positive test result is not required.” Note (a) just a ‘mention’ is sufficient and (b) the ‘mention’ doesn’t need a positive test result. These deaths are all covid deaths, sometimes called ‘involving covid’, of which by far the majority, typically around 90%, are also ‘due to covid’ ie covid was reckoned to be the underlying cause of death. A slightly longer definition from the monthly report might or might not make this clearer:

    “Definition of COVID-19: The doctor certifying a death can list all causes in the chain of events that led to the death and pre-existing conditions that may have contributed to the death. Using this information, we determine an underlying cause of death. We use the term “due to COVID-19” when referring only to deaths with an underlying cause of death of COVID-19. When taking into account all of the deaths that had COVID-19 mentioned anywhere on the death certificate, whether as an underlying cause or not, we use the term “involving COVID-19″. Age-standardised rates for deaths due to COVID-19 and involving COVID-19 are available in the accompanying dataset.

    Our definition of COVID-19 (regardless of whether it was the underlying cause or mentioned elsewhere on the death certificate) includes some cases where the certifying doctor suspected the death involved COVID-19 but was not certain. For example, a doctor may have clinically diagnosed COVID-19 based on symptoms, but this diagnosis may not have been confirmed because no test was available, or the test result was inconclusive.”

Leave a Comment

Your email address will not be published. Required fields are marked *