Thomas Galen (not his real name) took a degree in Medical Sciences and then spent thirty years working with models and numbers in finance and with large businesses.
Over the last few weeks, we have seen how hard it can be to understand how many people are dying and from what. For most people, this is a new experience, and perhaps somewhat baffling – they assume that death certificates are accurate and that we know how many people die and from what each year. The truth is, they are not and we don’t.
We now face an even bigger challenge. The government has chosen to shut down much of our economy based firstly on forecasts, but then on data on deaths from COVID-19. The problem is that the data is opaque and in some cases unreliable.
Dr. John Lee in the Spectator has given an excellent explanation of why making COVID-19 a notifiable disease has led to serious problems, not just in counting COVID-19 deaths, but in making comparisons with deaths from other diseases, notably flu.
To summarise his arguments, the problems with COVID-19 deaths are two-fold and are caused by the characteristics of those it kills – the elderly and those in ill-health. First, every day in England around 1,400 people die. In any epidemic, even one of a disease that kills no-one, some of those 1,400 will die infected by that disease. The percentage who do so will not be very different from the infection rate in the general population unless the disease infects certain age groups differentially, or we take some actions to protect certain age groups. With COVID-19, the evidence (e.g. the Diamond Princess cruise ship) suggests it infects adults at roughly the same rate regardless of age. Second, some people who die, die because they become infected by a virus or bacteria in what is inevitably the last week or so of their lives. What infection it is doesn’t usually matter much, and is not usually recorded. It is the infection that tips the balance, and contributes to the death, but is no more lethal than any other infection would be. If we test most people who die (particularly if we test post-mortem on a large scale), we are therefore – to some extent at least – simply testing the infection rate, not who is being killed by COVID-19.
What that suggests it that we are over-counting COVID-19 deaths. We can take a simple example of that to demonstrate the problem. The ONS breaks down deaths by location. Up to the week ending Friday 24th April (Week 17), it shows 300 cumulative COVID-19 deaths in hospices. That is clearly 300 too many. No doubt 300 people tested positive, but that was amongst people who were already terminally ill. It is quite wrong to count them as COVID-19 deaths. We can see that clearly, as the number of overall deaths in hospices over the period is exactly what we would expect from long term hospice averages.
The vast majority of recorded COVID-19 deaths are in hospitals (72%). There is no doubt that COVID-19 can make some people extremely ill and kill a small percentage of those. The evidence so far is that of those who die most have two or more other health issues. The question is whether they would have died if they had not contracted COVID-19? The answer in at least some instances, is yes. There is some evidence that some people dying in hospital of more obvious causes such as cancer, are being recorded as COVID-19 deaths because they have tested positive – because COVID-19, unlike flu and most other infections, is a notifiable disease.
For the elderly and frail a common problem caused by COVID-19 is pneumonia caused by a secondary infection. Would the patient have died without contracting bacterial pneumonia? We don’t know. Would they have contracted pneumonia without being infected with COVID-19? We don’t know. But we do know that many elderly and frail people contract pneumonia after getting infected with a whole range of other infections, not just COVID-19.
What that suggests is that a proportion of hospital COVID-19 deaths are not directly caused by COVID-19. It is impossible at this stage to know what proportion. We might have a better idea in a few months, when we can see more clearly how deaths in 2020 from other, non-COVID-19 causes compare to other years. At this point, all we can say that hospitals deaths are overstated, but we don’t know by how many.
This brings us to an interesting data point. The table below sets out the figures for deaths in hospitals from the ONS data:
|Average||Excess (number above average)||COVID-19 Deaths||Post COVID-19 Excess/(Deficit)|
If we assume that all COVID-19 deaths registered in hospitals are additional deaths, then we are ‘missing’ 5,600 deaths that we would normally see in hospitals. That could mean those deaths have been displaced to other places – homes and care homes – accounting for some of the excess deaths we see there. Alternatively, some of the COVID-19 deaths are deaths that would have happened anyway and so should not be counted as COVID 19. The truth is likely to be some mixture of the two.
If some deaths that would have occurred in hospitals have been displaced to homes and care homes because people are not going there then we have to consider that some people who would have usually been hospitalised and recovered, have instead not been hospitalised and died. That is a point we will come back to.
Are there any hospital deaths that are COVID-19 but are not recorded as such? It is possible, but it seems likely that there would be very few. For all the fuss about testing, the vast majority of tests (Pillar 1), even at the height of the epidemic, were negative – over 500,000 people. That means we are testing hundreds of thousands of people with symptoms that might be COVID-19, but which are not. We are also testing post-mortem, when there is a suspicion somebody might have died from COVID-19. Finally, we are counting some deaths as COVID-19 without testing because a doctor believes it might have been. It seems unlikely therefore that many COVID-19 deaths in hospitals are slipping through the net.
The other significant location (22%) of recorded COVID-19 deaths is care homes. What is happening with those deaths is somewhat murky. The figures are set out below:
|Cumulative 5 Year Average||Excess||COVID-19 Deaths||Post COVID-19 Excess|
Should we doubt those nearly 6,000 COVID-19 deaths? Probably. As far as we know, there is usually some time (days) from the onset of symptoms to death – COVID-19 doesn’t kill you quickly, let alone in your sleep overnight. If COVID-19 makes you ill enough to possibly die from it, people feel ill enough, and have enough time before they die to be hospitalised.
Moreover, given the age and health problems of those who have died in hospital, many must have come from care homes. Care homes are therefore well enough aware of the symptoms of COVID-19 and how important it is to send people to hospital. It seems unlikely that we have had nearly six thousand people (by 24th April) dying from COVID-19 in care homes but who were somehow not ill enough to be hospitalised prior to their death. Otherwise we must hypothesise that COVID-19 kills some of those in care homes more quickly but with fewer serious symptoms than others. That seems completely unlikely.
This strongly suggests that some, perhaps most, did not die of COVID-19, but with it. As a further piece of evidence, we now know that certification in care homes has become extremely lax. Care home deaths no longer need to be certified as COVID-19 by a doctor. For residents who have tested positive, why would care home directors not certify them as COVID-19 deaths? The ONS seems to be nervous about this and is treading carefully around care home deaths. It now says:
our regular weekly deaths release now provides a separate breakdown of the numbers of deaths involving COVID-19. That is, where COVID-19 or suspected COVID-19 was mentioned anywhere on the death certificate, including in combination with other health conditions.” (My emphasis).
These issues with care home deaths are backed up by testing. The CDC in the USA found 30% of care home residents tested positive, and similar testing in Belgium (a large sample) showed an infection rate of 20% in many care homes. The majority of those testing positive in both instances were asymptomatic in terms of COVID-19. Those that were symptomatic were not serious enough to be hospitalised. That last point is important – we can assume that the more serious cases from those care homes have already been hospitalised, and so we have higher rates of infection than measured. Cleary those tested in these care homes were not seriously ill and on the point of death. What that strongly suggests is that most of those who die in care homes are dying with the virus rather than from it. We can use the testing figures and say that 20-30% of those who die in care homes are dying infected rather than dying of COVID-19.
If we take the average number of deaths in care homes over seven weeks and use an infection rate of 25%, we get around 4,300 deaths. That represents those who would have died anyway but died infected. That compares with the COVID-19 total of 5,900 in care homes. We could therefore explain most of the care home COVID-19 deaths as simply being residents who have died after testing positive.
However, this leads to a different problem. We have a significant number of excess deaths in care homes even allowing for COVID-19. If we assume that 4,300 out of 5,900 of the deaths recorded as COVID-19 in care homes are not actually COVID-19, then we have a larger number of excess deaths. If we say only 1,600 are actually COVID-19, then we have total excess deaths of over 12,000. We can reduce that by subtracting some of our missing hospital deaths – say 2,500 – but that still means we have 9,500 people dying in care homes that we would not expect to see.
That sounds like a great deal, but we can in fact further reduce it. The cumulative deaths in 2020 up to Week 11 (the beginning of COVID-19) was around 2,000 fewer than the cumulative five year average. If we apply that deficit to the excess of deaths in care homes reduce the care home excess to 7,300. That number is clearly still large. Over however half can be accounted for by excess deaths amongst those over 90, and the rest by those over 85. It is possible that those deaths will reverse in the next few months, and we will have weeks with below average deaths in the very elderly. That would mean that we have brought forward deaths by a matter of weeks rather than years, perhaps because of the lockdown and the focus of the NHS on COVID-19.
Of note in this regard, PHE monitors respiratory infections across England. For Weeks 11 to 18, it has shown a very high number of non-COVID-19 infections in care homes – far above the usual number. For example, in their latest release PHE reported 648 acute respiratory outbreaks in care homes, but only 243 tested positive for COVID-19. Is there a second infection circulating in care homes that is responsible for the excess deaths amongst the elderly? We cannot know for sure, but it is possible.
We should also note that we appear to have displaced some deaths from hospitals to care homes and elsewhere. If that has happened, we have moved people who should be hospitalised out of hospitals. Many of those would have died anyway, no matter where they were, but it is surely impossible that we moved only those people, those who would have died, and not at least some people who would have been saved by being hospitalised? If that is the case, then some of the excess deaths are a direct consequence of the NHS’ COVID-19 policies and represent people the NHS would otherwise have saved but did not.
To summarise all of this:
|Average Deaths Weeks 11-17||
|Cumulative 2020 deficit to Week 10||
Note: These figures exclude deaths at locations other than hospitals and care homes. For the Unadjusted figures, that would give a further ~5,000 excess deaths.
The adjusted total for COVID-19 is 20,500, 75% of the unadjusted total. We could adjust that down further, particularly by further reducing COVID-19 deaths in care homes but at this stage it seems prudent to allow for some COVID-19 deaths outside hospitals. That may change as we understand more about the disease.
For the moment however, this is a reasonable readjustment to show we are significantly overcounting COVID-19 deaths and significantly undercounting excess non-COVID deaths. It is also reasonable to ask whether we have displaced unavoidable deaths from hospitals to other locations, and whether in doing so we have also displaced avoidable deaths and so contributed to the excess deaths elsewhere.