by Paul Bird
In assessing the impact of COVID-19 on the country, the most important statistic has been the number of deaths Covid has caused. That figure is relayed to us many times a day by the BBC and all other media.
But does it really convey the impact reliably, or are we being misled? I believe it is misleading. Firstly, there is some difficulty in defining the number itself. There have been many attempts, with definitions including: primary cause on the certificate; contributory cause on the certificate; death for any reason within 28 days or 60 days of having tested positive. None is regarded as completely satisfactory.
More misleading than that though is the question of interpreting the number. Does it tell us how much impact Covid is having on the UK population? How can you compare it with the impact of other diseases?
Crucially, the figure says nothing about how much life the victims would have had left to live if they had not died from the disease.
Quoting a number of deaths carries the implication that had the victims not succumbed to the disease then they would not have died. But we all have to die sometime. So a fatal Covid infection does not so much cause a death as bring it forward. Life is lost. In that sense, the death of a healthy 15 year-old who could expect to live another 65 years has more ‘impact’ than that of a 90 year-old with other advanced conditions from which he might expect to die sometime in the next month.
It is the ‘life lost’ rather than the occurrence of ‘a death’ that is important.
Figure 1 shows the average expectation of life in years for people by age in the UK based on the years 2017 to 2019.1 Obviously, Covid had no effect on these figures.
Figure 2 shows the numbers of deaths ‘involving’ Covid in 2020 in England and Wales, by age band.
I have multiplied the number of deaths in each age band by the average life expectancy for that band to give the number of years lost by all the people in that age band who died of Covid. So, for example, people aged between 70 and 74, had an average expectancy of life of 14.49 years in the years 2017 to 2019. Presumably, it would have been similar in 2020 had Covid not struck. We can therefore say that Covid deprived the 7,633 70-to-74 year-olds whose death involved Covid in 2020 of 110,602 (14.49 x 7,633) years of life.
Adding up the life-years lost in each band gives the total number of life-years lost to Covid in 2020 in England and Wales: 814,264.
How much is that?
It is difficult to relate to numbers like these. What do they mean?
Clearly, the overall size of the population is important. The impact of 100,000 people dying (for whatever reason) in a country of one million is rather more serious than in a country of 100 million.
The population of England and Wales in 2019 was just over 59 million, so 814,264 lost life-years is equivalent to 0.014 life-years per person, or five days.
Again, that figure is difficult to relate to. No-one wants to lose even one day of his or her life. However, the statistic does give the possibility of comparison with other things, such as the life lost to diseases and lifestyle choices we are more familiar with.
It also provides a way to gauge the proportionality of lockdown. How do the collateral harms of lockdown, which everyone is having to endure, compare to the life-time saved by lockdown? (At the time of writing it is not clear how many Covid deaths were saved by lockdown in 2020. Initial work on international comparisons suggests none at all.)
In strictly numerical terms, how does the time period five days compare to average life expectancy of 81.25 years?3 One way of visualising it would be to take a piece of string one metre long representing 81.25 years and cut off a piece representing the lifetime lost to Covid averaged over the whole population. How long would the offcut be? If you had been closely following the BBC’s coverage you might reason as follows: “Well, hundreds of thousands of people have lost their lives to Covid, and there are millions of people in the country, so maybe, I guess, the offcut would be a tenth? 10 centimetres?” That dedicated viewer would be wrong. The actual length of string corresponding to five days is 168 micrometres. That is, less than a fifth of a millimetre, a bit thicker than a human hair. You’d need specialist tooling to make the cut, and the string cut off would disintegrate into individual strands which would be hard to see with the naked eye.
In the above we used ‘nominal’ values.
Firstly, we used deaths ‘involving’ Covid, and secondly life expectancies for the average member of an age cohort. In my opinion those two values overstate the among of life lost to Covid in 2020.
Taking number of deaths first: there is confusion at the moment about how many people died primarily from Covid. As mentioned above, there are many definitions and none is really satisfactory. And more and more people are coming forward to complain about Covid being recorded as the cause of death on their relatives’ death certificates even though, in their view, it was not involved at all. The group Covid19 Assembly is carrying out a national audit of every official UK Covid death. I’d guess – there is no reliable data yet – that only three-quarters of the headline figure were actually deaths primarily from Covid. That assumption when applied to all age groups reduces the estimate of life-years lost in 2020 from 814,264 to 610,698.
Secondly, since almost all Covid deaths happen to people with one or more serious comorbidities one might assume their life expectancies would be shorter than the average for their age group.
For example, the average life expectancy for 82 year-olds is eight years (Figure 1), so that was used for that age band in my calculation above. Eighty-two is the average age of care home residents, but according to The British Geriatrics Society4 the average life expectancy in UK care homes is 24 months for care homes without nursing and 12 months for care homes with nursing.
I would have thought the same sort of argument would apply to all age groups. As a ‘what if’, not backed up yet by real data, let’s assume that the life expectancies of those who succumbed to Covid were half what would be expected for the average person in of their age. That is, for each age band you can halve the life expectancy in Figure 1.
That reduces the lifetime-lost estimate from 610,698 to 305,349 years.
As mentioned, those two modifications (the number of deaths and the life expectancies of those who died) to the official data aren’t based on any data, but I think they are reasonable and likely to be closer to the true picture which I hope will emerge when a more thorough analysis is eventually done.
305,349 years is equivalent to 0.005 life-years per person in England and Wales, or 1.9 days. That corresponds to 63 micrometres cut off our piece of string.
Comparisons with other diseases
We can apply the same methodology to other diseases, except that the life expectancy estimates in Figure 1 above will be slightly wrong. For Covid we can easily get life expectancies without Covid by looking at data from the years before Covid existed. We cannot do the same for things like heart disease or cancer, so the figures we have in Figure 1 include the mortality from the disease under consideration. An extra mathematical step is needed to correct for that. However, the effect is small if the loss of life from the disease under consideration is not too high a proportion of loss of life from all causes.
Firstly, cancer.4 That is, all cancers. Following the same procedure (but without the need to adjust for uncertainty in the real cause of death, or comorbidities) I get an annual life-lost figure of just over two million years. Nearly seven times my estimate for Covid.
Secondly, death from injuries, all types.5 705,087 life years lost in England and Wales in 2018, or 4.3 days per person. Over twice my estimate for Covid.
Thirdly, within the category ‘injuries’ are included suicides.6 195,910 life-years lost in England in 2019, or 1.3 days when divided by the population of England. 65% of the figure for Covid.
I have not got the figures for heart and circulatory diseases. I would guess they are of the same order as those for cancers.
The above tries to reveal a more accurate impression of the years of life lost to Covid than is given by simple death figures. But the impact of Covid, and of other diseases, is more than just time lost. Can the relative impact of premature death on an individual be measured solely in time left to live? What about the ‘quality’ of that time? Health economists regularly use the QALY concept – ‘Quality Adjusted Life Years’ – to account for that idea in planning resource allocation. A year of good health carries more weight than one spent with serious disability.
And what about the impact on others? Very sad though the death is of a frail 90 year-old in a care home who has had a full and satisfying life, it cannot be as tragic as that of, say, her 35 year-old granddaughter with a husband and two children who might have died prematurely of breast cancer due to lack of diagnosis and treatment in lockdown.
Both of those considerations suggest society’s assessment of Covid harms has been exaggerated compared with those of other more familiar diseases, and with the harms of Lockdown.
- I have argued that the crude death statistics we have relied on to assess the damage done by COVID-19 are misleading.
- A different and more meaningful metric is proposed: the sum of life-years lost by all those who died in a particular period divided by the whole population.
- Comparing that figure for Covid in 2020 with those for other causes of death suggests that Covid had a much lower impact than the the impression created by the mainstream media: around a seventh that of cancer, a similar fraction for heart disease, and a half that of physical injury.
Paul Bird is an Engineering Consultant.
7 Time for a smoke? One cigarette reduces your life by 11 minutes M Shaw, R Mitchell and D Dorling, University of Bristol. BMJ 2000 Jan 1; 320(7226): 53.