Give us the facts

Over the past year (essentially the ‘Covid’ year) I’ve been looking at death and population data (for England and Wales in particular). When I was young I certainly never thought that would be my hobby as I approached normal retirement age; we live in strange times.

One thing that has struck me is the unreliability of the information supplied as ‘data’ by various agencies and the uncritical way most media report on these data.

I offer as an example the definition of a ‘case’ of Covid-19 being asserted by a positive PCR test. This ignores the fact that the ‘D’ in Covid-19 stands for ‘disease’; that is to say ‘dis-ease’, a state of being unwell. Public health messaging (in England) regularly reminds us that one in three people who test positive have no symptoms at all. In other words they are not dis-eased at all. Whatever. That ship sailed a while ago when the World Health Organisation also adopted this ridiculous definition in December 2020. The result is that one of the crucial numbers in assessing the epidemic in England and Wales (the number of ‘cases’) is almost useless.

Let us turn our attention partially away from the damned Coronavirus bug and consider what the majority of people die of: I mean, ‘getting old’.

The Office of National Statistics has regularly published datasets since 2010 of how many people die each week (actually, how many deaths are registered, which is slightly different). Until 2020 these have been divided into how many people in various sex and age groups (M/F, ‘Under 1’, 1-14, 15-44, 45-64, 65-74, 75-84, 85+) die each week. Since 2020, they’ve subdivided the data into narrower bands (‘under 1’, ‘1-4’, ‘5-9’ and then 5-year bands up to ‘85-89’ and finishing with ‘90+’). These smaller divisions may be useful for comparisons in future years and can be easily combined to compare against previous years’ data.

• Simple assessment of the data shows that about half a million people die in England and Wales each year. Sometimes more, sometimes fewer.

• Annual overall deaths are about 50:50 women:men.

• A lot more old people die than young people. Not too surprising really.

• Fewer people die each week in the Summer months* than in the Winter months. Some people speculate that the warmer weather or longer days somehow help people to live longer; perhaps Vitamin D (the human skin makes Vit D when exposed to sunlight) is involved?

• The seasonal variation in death numbers affects revealed in the weekly death registration count affect older age groups more than younger ones. There’s almost no seasonal variation in death registration rates for the under 65’s (ie no Winter spike).

• A slightly closer assessment shows that men tend to die younger than women - which has the inevitable effect of there being considerably more women than men among the elderly.

• Because more women than men reach the age of 85+ more women than men in this age group die. No savage indictment of current or previous government policy is required as a response; but if you find it helps you, then feel free to let rip.

*Some may be surprised at considering the seasonal variation as fewer people dying in Summer rather than more people dying in Winter. The key point is that everyone will die eventually; and about 1% of the World’s population die each year. It is therefore appropriate to consider living people to be not dead yet.

In addition to the weekly data, ONS publish an annual report counting the number of deaths in each group for each of a number of causes of death (these are derived from the main cause of death recorded on the death certificates). We can access the data for 2013-2020 through Nomisweb. Causes of death are coded using the ICD-10 codes (for example, if someone’s death is recorded as ‘Myocardial Infarction’ (a type of heart attack) then ONS would probably code the death as ‘I21’ or one of the subcodes). The annual data for England & Wales is sometimes accessible as early as August the following year, so the data for 2021 may be available in mid-2022.

The weekly datasets do not include full ‘cause of death’ data. However they have historically included data on ‘Deaths where the underlying cause was respiratory disease (ICD-10 J00-J99)’ and in 2020 they included ‘Deaths where COVID-19 was mentioned on the death certificate (ICD-10 U07.1 and U07.2)’. By comparing the annual dataset for 2020 with the combined weekly datasets we find that the total death registrations where Covid was the underlying (ie main) cause of death are about 80% of the registrations where Covid was mentioned on the death certificate.

ICD-10 codes are more generally used for the analysis of health data and as a result, some codes just do not appear in death stats (for example there is no recorded death coded as H93: ‘Other disorders of ear, not elsewhere classified’, including H93.1: ‘Tinitis’). Other codes are almost exclusively used for death diagnosis in the elderly such as R54: ‘Senility’ (this refers to ‘old age’, not mental faculties).

• It should not surprise you to find that nobody under 70 years has R54: ‘Senility’ (Old Age) recorded on their death certificate. If a doctor put that on a certificate for someone under 70 the death might well be referred to a Coroner’s court for investigation and a proper diagnosis.

• Main cause of death (ONS slightly confusingly calls this ‘underlying’ cause of death) is very much a matter of opinion by the certifying doctor.

• Many ICD-10 codes are grouped together into 47 ‘Leading Causes’ of death. These ‘Leading Causes’ do not cover all possible codes so we’re left with a 48th group which I’ve labelled ‘Other’

• In 2020 607,099 deaths were registered in England and Wales. It won’t surprise you that is higher than usual. This was made up of 307,572 males and 299,527 females (50.7:49.3).

• 15,710 (2.59%) death registrations were attributed to ‘LC01 Accidents’

• 677 (0.11%) death registrations were attributed to ‘LC25 Homicide and probable homicide’

• 5,203 (0.86%) death registrations were attributed to ‘LC41 Suicide and injury/poisoning of undetermined intent’

• The death registrations excluding Accident, Homicide and Suicide could be considered to be ‘Natural Causes’. However, bear in mind that some suicides may be deaths of people with unbearable health conditions who might have been expected to die ‘naturally’ soon.

• 73,681 (12.14%) deaths were attributed to ‘LC47 Covid-19’. This number is lower than other Covid-19 death figures published in the ONS weekly stats or PHE or UK Government dashboard figures.

• There were 533,418 death registrations (607,099 - 73,681) where Covid was not shown as the cause of death.

If we chart the number of natural death registrations (ie excluding Accident, Homicide, Suicide) in England and Wales in 2013 to 2020 against age group and sex we see this pattern:

Surely there’s something wrong here? The female chart scale goes up to 90,000 deaths but the male chart scale only goes up to 60,000 deaths. But didn’t we show above that male:female deaths are about 50:50? Also, in every year the number of males dying declines over age 85. What’s going on?

The answer is, of course, that fewer males than females survive to age 85+ and because there are fewer of them, there are fewer deaths. If instead of death registrations we chart death rates as a percentage of each sex and age group we gain a better understanding:

That makes more sense. The male chart scale goes up to 30% and the female chart scale goes up to 25% and the death rates keep increasing with age.

Comparing the years 2013-2020: clearly 2020 shows a spike relative to the other years and that’s obviously a bad thing (Covid) - but I don’t think the relative height of the 2020 columns is frighteningly higher than the 2015 columns (2014/15 was a bad ‘flu season). Also, note that the 2019 columns show a distinct ‘dip’ relative to the others. Fewer people than usual died in 2019 which may mean that the population in 2020 had more frail people than usual.

We can simplify the charts further by averaging the years 2013-2019 so we can easily compare the recent pre-Covid years with the Covid year. We can also combine the male and female charts to better highlight the difference between the sexes.

All the data series seem to show a neat progression of increasing death rate with advancing age (apart from the under 10s, the percentage death rate increases by a factor of about 1.8 times for each additional five years of age; further calculation will reveal that the death risk approximately doubles for every 7 years advance in age - which means that for 70 years advance in age (90 compared to 20 years old) the risk of natural causes death is about 1,024 (2^10) times higher. If we use a logarithmic scale for the death rate axis we reveal this remarkable pattern:

That’s a very clear association between age and death rate. If we calculate the best fit curves for males and females in the 10 to 85 age range (ie excluding the under 10s and the catch-all ‘over 90s’ groups) we find that for the 2013-19 average and in 2020 life is 100% fatal for males in a projected 105-109 age group and females at around 110-114 years. (NB this is statistics; there may be one or two exceptions to this rule). The key point is that with or without Covid the fact is that the older we are, the more likely it is we will die this year.

NB. Be aware that a logarithmic scale cannot start at zero. In this instance the y-axis starts at a rate of 0.001% which represents about 18 5-9 year olds (the population group with the lowest death rate).

Don’t run away with the idea that there’s no difference in death rate between 2020 and the 2013-2019 average. The difference is still there but the logarithmic scale makes it slightly more difficult to see on the chart. It still represents about 50,000 extra deaths (an extra 0.077% risk across the whole population) - but not the crazy high numbers predicted elsewhere. What this chart shows is a consistent strong relationship between age and susceptibility to death by natural causes.

At this point you may be wondering whether Covid-19 deaths are as strongly age associated or if they’re just ‘extra’ to the normal annual death toll. Wonder no more.

Death from Covid-19 is also very strongly age associated. NB the logarithmic scale in this instance starts at 0.0004%.

As at 12 July 2021 (this is early in the year compared with previous years) ONS have published the data for causes of death (as reported on death certificates) for England and Wales in 2020.

The data is made available summarised in a number of ways and I have chosen to download and present data about deaths assigned to the 47 Leading Causes of death (LC01 - LC47) expressed as a percentage of age group and sex. The ‘Leading Causes’ do not include all causes and so we have an additional category presented as ‘Other’.

So what does it look like?

If you read the chart and think that it shows that about 25% of males aged 90+ died in 2020, then: Congratulations! That is what it shows. The figure is actually about 25.8%. It’s not that shocking; the average death rate for males 90 years and over in 2013-2019 was about 22%. Yes, I’m saying that out of a population of roughly 170,000 men aged 90+ in England and Wales 37,000 dying each year is normal. It also means that each year about 37,000 men survive being 89 to ‘graduate’ into the 90+ group.

Look at it this way: if you’re a male of average health aged 90 you’re 3 times more likely to survive the year than not. So what do you think you should spend most of your time on?

Well, I don’t know about you, but at a scale suitable for the age 90+ columns I can’t read anything in the columns below about age 60-64. What we’ll have to do is use different scales for the different age groups. Clicking the image below will start a slideshow (it’s probably a good idea to open it in a new window) with a scale appropriate for the lowest death rate (among 5-9 year olds; males 0.007%, females 0.004%) and gradually increase the scale and introduce additional columns when they fit.

Points of interest:

• among 5-9 year olds, most deaths are attributed to ‘Other’ ie not to one of the ‘Leading Causes’. The most significant of the identified causes is LC02: ‘Cancer (malignant neoplasms)’.
• 10-14 year olds died of LC02 at a similar rate to the 5-9s but shockingly we see some LC41: ‘Suicide and injury/poisoning of undetermined intent’ make up some of the total.
• among 1-4 year olds, we see a similar rate for LC02 but in addition, LC01: ‘Accidents’, LC17: ‘Congenital malformations, deformations and chromosomal abnormalities’ and LC42: ‘Symptoms, signs and ill-defined conditions’.
  • LC42 can be loosely characterised as ‘we don’t really know exactly why, but the person died’.
  • The ‘Other’ category means that ‘We do know why this person died - but it’s not in one of the 47 leading causes’.
• 15-19 year olds died at a much higher rate than the 10-14s. In this column we see a distressingly large relative increase in LC25: ‘Homicide and probable homicide’.
• The 20-24 year olds column shows the first visible impact of LC47: ‘COVID-19’. We also see a significant increase in deaths of males (compared with females of a similar age). The increases are predominantly in LC01 and LC41 (accidents and suicide).
• The previous trend continues for 25-29, 30-34 and 35-39 year olds.
• The death rates climb as age increases but at around the same general level as 50-54 year olds we also see the introduction of a column for the under 1 year olds at about 0.41% of males and 0.34% of females dying in their first year.
• The trend continues with a larger proportion of people dying in each step up the age groups until in the 60-64 age group (my group) about 1% of us males die each year; 0.65% of females.
• The trend continues with advancing age but causes of death begin to include LC18: ‘Dementia and Alzheimer disease’ and also LC42: ‘Symptoms, signs and ill-defined conditions’ makes a reappearance. Among the elderly LC42 usually means the death certificate lists ‘Senility’ or ‘Old age’ as the cause of death.
• The trend for adults (say, over 20s) excluding death due to LC01, LC25 and LC41 (accident, homicide, suicide) closely follows a mathematical curve which can be extrapolated to reach a maximum of 100% (everyone dead) at about age 105 for males or 110 for females. Again, not too surprising; very few people live that long.

• For males 10-49 and females 10-39 LC41: ‘Suicide and injury/poisoning of undetermined intent’ accounted for more deaths than LC47: ‘COVID-19’. (Note that the LC41 rate did not significantly increase in 2020 compared with previous years.)

• For males and females under 50 years LC01: ‘Accidents’ accounted for more deaths than LC47: ‘COVID-19’.
• For males under 35 and females under 20 LC25: ‘Homicide and probable homicide’ (ie murder and manslaughter) accounted for more deaths than LC47.
• For males 25-55 and females 30-55 LC16: ‘Cirrhosis and other diseases of liver’ resulted in more deaths than LC47.
• For every group except females over 90, deaths from LC02: ‘Cancer (malignant neoplasms)’ exceeded deaths from LC47.
• LC18: ‘Dementia and Alzheimer disease’ killed more males over 90 and females over 80 than LC47. I don’t know about you, but I fear dying of Alzheimer’s more than Covid-19

It seems to me that we (the world, not just the UK or England and Wales) should be spending more to address deaths from each of Accidents, Homicide and Suicide than we spent on Covid-19 in 2020.

On 22nd March 2020 SAGE minutes included this: Options for increasing adherence to social distancing measures.

In the paper the advisers offer:

Perceived threat: A substantial number of people still do not feel sufficiently personally threatened; it could be that they are reassured by the low death rate in their demographic group, although levels of concern may be rising. Having a good understanding of the risk has been found to be positively associated with adoption of COVID-19 social distancing measures in Hong Kong. The perceived level of personal threat needs to be increased among those who are complacent, using hard-hitting emotional messaging. To be effective this must also empower people by making clear the actions they can take to reduce the threat.

• It appears that they didn’t think a good understanding of the risk should include comparing the threat of Covid-19 to the threat of accidents or many other leading causes of death.
• The phrase ‘hard-hitting emotional messaging’ suggests to me that they proposed scaring people rather than educating us.
• ‘To be effective this must also empower people by making clear the actions they can take to reduce the threat’. So:
  • first we must increase the perception of threat
  • then tell people how (in our opinion) to avoid the threat of this new bug
• You did spot the phrase ‘they are reassured by the low death rate in their demographic group’ - didn’t you?

To be fair, we have the benefit of hindsight and access to the information of what was actually diagnosed throughout 2020. The government advisers were offering their thougths just before the first UK ‘lockdown’ was imposed and amid reports from China that the lockdown in Wuhan had been a complete success. In addition, Report 9 had just been published by Prof Neil Ferguson’s group at Imperial College London (on behalf of the ‘Imperial College COVID-19 Response Team’) in which they assess that:

In the (unlikely) absence of any control measures or spontaneous changes in individual behaviour, we would expect a peak in mortality (daily deaths) to occur after approximately 3 months (Figure 1A). In such scenarios, given an estimated R 0 of 2.4, we predict 81% of the GB and US populations would be infected over the course of the epidemic. Epidemic timings are approximate given the limitations of surveillance data in both countries: The epidemic is predicted to be broader in the US than in GB and to peak slightly later. This is due to the larger geographic scale of the US, resulting in more distinct localised epidemics across states (Figure 1B) than seen across GB. The higher peak in mortality in GB is due to the smaller size of the country and its older population compared with the US. In total, in an unmitigated epidemic, we would predict approximately 510,000 deaths in GB and 2.2 million in the US, not accounting for the potential negative effects of health systems being overwhelmed on mortality.

Note that the prediction of 510,000 deaths is for ‘GB’ ie England, Wales and Scotland. A follow up report also from Prof Ferguson and many of the contributors to ‘Report 9’ (including the ‘Imperial College COVID-19 Response Team’) in early June effectively argued that because deaths had not reached the predicted number, the interventions made must have been successful - this is a circular argument. It also extolled the Chinese response:

In China, strict movement restrictions and other measures (including case isolation and quarantine) began to be introduced from 23 January 2020, which achieved a downward trend in the number of confirmed new cases during February and resulted in zero new confirmed indigenous cases in Wuhan by 19 March 2020.

I don’t think the link between cause (strict movement restrictions) and supposed effect (zero new confirmed indigenous cases) has been demonstrated. I also think the Chinese authorities have a poor track record when it comes to telling the truth about the health and wellbeing of The People. This BBC article from 7 Feb 2020 is a good example:

[Dr] Li Wenliang died after contracting the virus while treating patients in Wuhan.

Last December he sent a message to fellow medics warning of a virus he thought looked like Sars - another deadly coronavirus.

But he was told by police to “stop making false comments” and was investigated for “spreading rumours”.

and this:

China’s leadership had already faced accusations of downplaying the severity of the virus - and initially trying to keep it secret.

The government has admitted “shortcomings and deficiencies” in its response to the virus, which has now killed 636 people and infected 31,198 in mainland China.

and this:

And there is still a massive mismatch between the figures released by China and what outbreak analysts think is really going on.

It is widely accepted that the laboratory-confirmed cases are just the tip of the iceberg.

Some mathematic models of the outbreak suggest the true size of the epidemic could be 10 times higher than the official figures.

…and in this article from 28 Dec 2020:

A Chinese citizen journalist who covered Wuhan’s coronavirus outbreak has been jailed for four years.

Zhang Zhan was found guilty of “picking quarrels and provoking trouble”, a frequent charge against activists.

Yet somehow we’re supposed to believe them when they tell us that the lockdown in Wuhan worked?

According to Worldometers (who draw their data from the various countries’ official figures) The People’s Republic of China has had only 4 Covid-19 deaths since 18 April 2020 - no, I don’t believe that either; it’s a blatant lie. Not only do the Chinese authorities not tell the truth about such matters, they don’t really care if you believe them or not; they don’t even try to make their reports convincing. It astonishes me that the Imperial College Covid-19 Response Team extolled the Chinese figures in their report.

To summarise: I think China has grossly understated the number of Covid deaths and grossly overstated the effectiveness of their lockdown policy.

Let us consider the prediction of 510,000 deaths in GB from Covid-19, excluding any additional deaths due to swamping the NHS:

• Various people attempt to defend ‘Report 9’ by stating that it was conditional and not a ‘prediction’. In the extract paragraph the word predict or predicted is used 3 times.
• It’s interesting that the phrase ‘we predict’ is used first followed by ‘is predicted’ and later ‘we would predict’. It sounds to me like they were increasingly unsure of themselves.
• I would summarise the extract paragraph as ‘if all our assumptions are correct then we predict this result’ - but the assumptions are notably bogus. For example, ‘spontaneous changes in individual behaviour’ do occur when there’s a nasty ‘flu going around. A study concerning the 1918 ‘flu epidemics published in 2007 stated: ‘Our analysis also suggests that individuals reactively reduced their contact rates in response to high levels of mortality during the pandemic.’; The authors? Oh yes, Martin C. J. Bootsma and Neil M. Ferguson. And in this study by Jude Bayham et al from 2015: ‘We provide empirical evidence that Americans voluntarily reduced their time spent in public places during the 2009 A/H1N1 swine flu, and that these behavioural shifts were of a magnitude capable of reducing the total number of cases.’
• Elsewhere that Report 9 paragraph states ‘Epidemic timings are approximate given the limitations of surveillance data in both countries’. How can the predicted timings be ‘approximate’? I suppose they mean that the confidence interval is such a wide range as to be useless for planning. A better word would be ‘unreliable’ - as in ‘do not rely on this’.
• The report predicts a doubling of the usual annual death rate - we typically see around half a million deaths in England and Wales each year.
• If the epidemic followed a normal epidemic (Gompertz) curve (the figure 1A referred to in the extract paragraph appears to correspond roughly to a Gompertz curve), the peak rate would occur when roughly 1/3rd of the eventual deaths had occurred (ie the rate of death would begin to slow down after about 170,000 of the predicted 510,000 deaths).
  • The actual peak rate of the so-called ‘first wave’ occurred in week 16 of 2020 (week ending 17 April 2020) when 8,758 Covid-19 death registrations were recorded - just six weeks after the first reported death registrations in England and Wales, and after about 19,000 deaths had been attributed to Covid-19 (an order of magnitude below the Imperial College COVID-19 Response Team prediction).
  • The prediction was for the rate to keep increasing for 3 months - ‘approximately’.

Some would continue to argue that Covid-19 deaths would have been higher without the lockdowns and other restrictions - but I’ve seen no convincing evidence to support that position.