Putting a bionic cat among the catatonic pigeons, the much publicised only 17k deaths from covid figure has led to some of the most acrimonious splats Dr No has ever seen on squitter, the social media platform on which participants dump on each other from great heights. Joining the general opprobrium, Tim Harford used the first slot on this week’s More or Less programme on Radio 4 to apply ‘tireless debunkery’ to the claim, with more or less help from a molecular biologist who got her ONS numbers mixed up, claiming (at about 3:12) that the 17k deaths were covid–19 deaths where “no other health condition was mentioned on the death certificate”, which is not true: the actual number of deaths where no other condition was mentioned — that is, deaths where covid–19 was the only cause mentioned — is even smaller, at 6,183. The 17k figure, on the other hand, represents covid–19 deaths with no pre-existing conditions, or as ONS also calls them, deaths from covid-19 with no other underlying causes; but that is not to say there were no other causes mentioned. Other causes may have been present, but they were not deemed by ONS to be pre-existing, or underlying causes.
This matters, because of the way ONS codes the underlying cause of death when covid–19 appears in Part I of the MCCD, or Medical Certificate of Cause of Death. Undeterred, the radio presenter with a voice like audio Horlicks ploughed on: as it is with hand grenades and horseshoes, in the land of tireless More or Less debunkery, near enough is more or less good enough. After some preliminary waffle, the piece then degenerated into a caring and sharing so much it hurts wokist wail about not counting the tens of thousands of alleged additional covid–19 deaths, over and above the 17k deaths, that appear, for example, on the government’s coronavirus dashboard. No one, least of all Dr No, is not counting these deaths. Instead, all he wants to know is where these additional deaths should be counted: in a covid–19 column, or in a non-covid–19 column?
To understand what is going on here, we need to return again to how ONS codes causes of death. The covid–19 coding rules come from WHO, and they place a premium on collecting as many deaths with covid–19 as underlying cause deaths as possible, reflecting not medical veracity, but what WHO airily calls “interests of importance for public health”. The rules are crystal clear that public interests take precedence over medical accuracy: “…whether a sequence [chain of events leading to death, and so underlying cause] is listed as ‘rejected’ or ‘accepted’ may reflect interests of importance for public health rather than what is acceptable from a purely medical point of view. Therefore, always apply these instructions, whether they can be considered medically correct or not (emphasis added).”
These instructions, which apply to a number of infectious and parasitic diseases, are in fact an established pre-covid–19 procedure, embedded in WHO’s ICD–10 Volume 2 Instruction Manual Fifth Edition dated 2016, which has been adopted by ONS, and indeed most other nations. Coders are instructed to reject sequences of events leading to death that place these diseases as due to any other causes (ie they are not the underlying cause): “Do not accept [these] infectious and parasitic diseases as due to any other causes, not even HIV/AIDS, malignant neoplasms or conditions impairing the immune system”. The effect is that these infectious and parasitic diseases almost always (there are a few exceptions, for instance external causes such as trauma) trump any other potentially valid underlying cause of death to become the underlying cause of death, even if it makes no sense from a “purely medical point of view”. All that happened when covid–19 arrived was that these rules were also applied to covid–19.
The effect of this rule, as noted, is that the infectious diseases listed, including covid–19, almost always trump other potentially valid underlying causes of death. A sequence of events such as metastatic lung cancer leading to covid–19 infection leading to pneumonia will be coded with covid–19 as the underlying cause, when in non-covid–19 circumstances lung cancer would be the underlying cause. Likewise, the normally accepted sequence Alzheimer’s disease leading to immobility and wasting leading to bronchopneumonia, with Alzheimer’s as underlying cause, would, if covid–19 was present on a higher line — as shown in the image at the top of this post — be coded with covid–19 as the underlying cause, even if it makes no sense from a “purely medical point of view”. We can all no doubt see where WHO, and all the organisations that have adopted WHO’s guidelines, are coming from: an ardent desire to bump up the numbers to grab global and political attention in the face of certain infectious diseases. But at the same time, we should not forget that, by the same token, covid–19 death counts are deliberate collected in a manner intended to “reflect interests of importance for public health…whether medically correct or not”. Or to put it more bluntly: the numbers may be inflated, possibly wildly inflated.
We can now return to the 6,183 deaths, the 17k deaths, and more or less 140k deaths — all the counts cover, give or take a few days, from the start of the pandemic to the end of 2021 — and consider what they really represent. Harford’s stooge may have got her knickers in a twist, but we can straighten them out. The 6,183 deaths are those where COVID-19 was the only cause mentioned. Dr No takes this to mean there was only one entry in Part I of the MCCD (Medical Certificate of Cause Death), and it said covid–19: the underlying cause of death was covid–19, the whole covid, and nothing but the covid. Fair enough: if there is only one entry, then that has — unless it is nonesense — to be the underlying cause of death. But that only covers a tiny fraction, 6,183, of all the deaths. Nonetheless, these 6,183 deaths were true (insofar as the MCCD records the death) covid–19 deaths.
The 17k covid–19 deaths — often this FOI response is often given as the source, but more up to date and detailed routine data that was used for the FOI response can be found here and it includes data to the end of 2021 — are harder to tease out. The count is actually a count of all covid deaths with no pre-existing conditions, which, as we noted earlier, is not the same thing as no other conditions. There may well have been other conditions present in Part I of the MCCD, but, for whatever reason, they were not deemed to be pre-existing, or underlying cause of death, conditions. And herein lies the rub: there is no way of knowing how often in these deaths covid–19 made sense “from a purely medical point of view” — ie covid–19 really was the true underlying cause of death — and how often, under the covid–19 coding rules, covid–19 trumped another perfectly legitimate underlying cause of death that, without covid–19 presence on the MCCD, would undoubtedly be the underlying cause of death. Even worse, exactly the same problem applies to all the other “due to covid–19” deaths, all 140k of them (140k being 175k less 10% as covid–19 was not the underlying cause, even though it was on the MCCD, and then minus the 17k).
Dr No suggests that from a true public health perspective, this ambiguity in the numbers is a epidemiological disaster, which means nothing less than we have no accurate way of knowing the true extent of covid–19 mortality. Is the sequence of events shown in the image at the top of the post — Alzheimer’s disease leading to Immobility and wasting leading to Covid–19 — really a covid–19 death, as ONS would count it, or it really an Alzheimer’s death, with covid–19 being the final consequent event in the sequence of events that lead to death? How many times has covid–19 been allowed to trump an otherwise perfectly valid underlying cause of death, that indeed would have been the cause of death, had covid–19 not appeared on Part I of the MCCD?
Even if they have the data, which Dr No doubts, ONS are hardly likely to tell us how many times covid–19 was allowed to trump another valid underlying cause. Dr No’s favourite statistic, all cause mortality, is of no help on this, because there is no cause of death information in the data, and so, if there are excess deaths, we cannot even know whether they are covid–19 deaths, or covid–19 response deaths, caused by lockdowns, delayed diagnosis and treatment, let alone whether covid–19 trumped other causes. But there is one thing we can look at that might give us a somewhat vague clue — do not for a moment consider it more than that — as to how often covid–19 steals the underlying cause of death from the jaws of other normally common underlying causes of death: the numbers of those deaths from common causes in 2020 and 2021, compared to those in previous years. Our hypothesis, such as it is, is that if covid-19 has trumped deaths from these other causes, then there will less of them in the last two years.
Dr No cannot underline strongly enough that this is a very oblique way of approaching this question. It may be that covid–19 trumping, when it happened, was legitimate: covid–19 really was the underlying cause of death. But then again, maybe not; and as we have no other way of approaching the question, it is at least a way of looking at the data, that might give us cause for thought. With the caveats established, Table 1 shows the average annual numbers of deaths for common causes for 2015-2019, and the numbers in 2020 and 2021. We’ll stick with numbers, which we have covered so far in this post, and which are easier to grasp than the also available but more abstract age standardised rates, though they tell pretty much the same story. The data comes from ONS’s monthly mortality reports, available here.
|Underlying Cause of Death||2015-2019||2020||2021|
|Dementia and Alzheimer's disease||61,928||66,060||57,673|
|Ischaemic heart diseases||53,429||51,979||52,821|
|Malignant neoplasm of trachea, bronchus and lung||28,108||26,571||26,389|
|Chronic lower respiratory diseases||29,681||26,917||24,319|
|Symptoms, signs and ill-defined conditions||12,078||14,385||15,242|
|Influenza and pneumonia||25,969||18,656||15,148|
|Malignant neoplasm of colon, sigmoid, rectum and anus||13,866||15,960||14,266|
Table 1: number of deaths for leading causes of deaths, 2015-2019 (average number) and 2020 (number) and 2021 (number)
The numbers for many, but not all, causes of death fell in 2020 and 2021, compared to the previous five years. In most cases the falls were small, 10% or less, but for influenza and pneumonia, perhaps the most likely candidates to be trumped by covid, the falls were 28% (2020) and 42% (2021). Curiously, the numbers of lower bowel cancer deaths increased in both 2020 and 2021 compared to the previous five years, as did deaths from symptoms, signs and ill-defined conditions. Overall, there were 6,793 less deaths from these causes in 2020, and 20,052 less deaths in 2021, compared to the previous five years. Dr No suggests this data doesn’t rule out covid–19 trumping other causes, but at the same time, neither does it show beyond doubt that trumping happened. All in all, rather a more or less result…
Where does this leave us? We have 6,183 deaths that can be considered true covid deaths, at least by the lights of the entry on the MCCD. We have around 11k (17k minus 6k, because the 6k will be in the 17k) deaths where there may have been other entries on the part I, but if present, they were discarded, and around 140k deaths that appear to have had pre-existing conditions — defined by ONS as “the last health condition mentioned on the first part of the death certificate (the direct sequence of events leading to death) when it is recorded on a lower line to, and therefore clearly preceding, the coronavirus (COVID-19); and all mentions in the second part” — and yet covid–19 was recorded as the underlying cause of death. We still have no way of knowing how often the trumping of the underlying cause of death by covid–19 was legitimate, and how often it was done to reflect “interests of importance for public health rather than what is acceptable from a purely medical point of view”.
What can we learn from this sorry state of affairs? Clearly, very little about how many people really died due to covid—19, because the covid trumping rules so muddy the waters that is impossible to disentangle real due to covid–19 deaths from those hoovered up by the covid–19 trumping rules. The possible range runs, more or less, from 6,183 to 140k, and the failure to have not the slightest clue of where the real number lies represents a colossal failure of public health medicine. Perhaps that leads to the real lesson: in the face of a pandemic of an infectious disease that comes under the special coding rules, you cannot trust the numbers of deaths to be medically correct, because the numbers are not even meant to be medically correct; instead, they are produced to reflect those mysterious interests of importance for public health — whatever they may be.
Edit 1600 29 Jan 2022: inverted percentages in paragraph under Table 1 corrected (72% to 28%, 58% to 42%).