The UK death figures released by the Office of National Statistics for weeks 14 to 16 indicate that there are excess deaths over the five-year average that are not accounted for by the official number of ‘with Covid-19’ deaths. Are these excess deaths undiagnosed Covid-19 deaths, or deaths caused by the lockdown?

Dr Malcom Kendrick has written about how he has had to sign more death certificates than usual in care homes, often with no way of checking whether the patients had Covid-19, so Covid-19 has not gone on to these death certificates even though he suspected that some of these patients may have had it. This suggests that there may be undiagnosed Covid-19 deaths occurring, especially in care homes.

But is this the explanation of all the excess deaths? Are we also seeing some excess non-Covid deaths that are being caused by the lockdown itself, and by the associated media panic? We have some general grounds to suspect this, seeing as so many senior medics, such as Dr Katherine Henderson, the president of the Royal College of Emergency Medicine (which represents A&E doctors), Prof Martin Marshall, the chair of the Royal College of GPs, and Dr Simon Walsh, the British Medical Association’s lead for emergency care, have issued desperate pleas for people who are seriously ill to continue to use the emergency services and hospitals. In addition, there are anecdotal reports from ambulance workers who say they are arriving at homes to find heart attack patients already dead much more often than usual.

A closer analysis of the ONS numbers supports the view that there really are substantial numbers of people being killed by the lockdown.

First of all, how many excess non-Covid deaths do we have? Let’s consider week 16, the week ending 17 April. The ONS reported 22 351 all-cause deaths in England and Wales. (Note that the ONS spreadsheets are organised by the date the death certificates come in, not by day of death.) This was 11 854 more than the five year average for week 16. But there were only 8758 Covid-19-related deaths in week 16. That’s 3096 excess deaths in week 16 not accounted for by the Covid-19 numbers. (In week 15, the week ending April 10, there were 1783 excess non-Covid deaths, and in week 14, the week ending April 3, there were 2607.)

Note that the Covid-19 numbers include anyone who died with Covid-19, not necessarily of it, so in reality there will probably be fewer people who died of Covid-19, but I’m going ignore that complication here. Ignoring that makes my case harder to prove, so I’m not gaining an advantage by doing it.

There’s no direct way of determining whether or not these excess non-Covid deaths really were Covid cases that failed to be diagnosed. However, there is one thing we know about Covid — it kills a lot more men than women. In week 16, for example, there were 8758 Covid-related deaths, and that broke down by sex as follows: 5057 males, and 3701 females. Percentage-wise that’s around 58 to 42 (57.74 to 42.25). For weeks 14 and 15 the percentage difference was even greater, around 61-62 to 38-39. These figures are all in line with what has been seen around the world for Covid-19. The WHO, for instance, says 63% of all ‘with Covid-19’ deaths are male.

If our 3096 excess deaths for week 16 were all undiagnosed Covid deaths, then we would expect that to break down as follows (using the percentages 57.74 for men and 42.25 for women):

Male: 1788 deaths

Female: 1308 deaths

Difference: 480.

We’d expect to see the remaining 10 497 non-Covid deaths (the deaths you normally get in week 16) to follow the usual sex differences. Here we face a tricky issue, though: what are the usual sex differences in regards to mortality in England and Wales? They are not, as some might suppose, 50:50. Normally in England and Wales a few more women die than men each week.

The ONS has a spreadsheet showing annual deaths up to 2018, with sex breakdowns since 1956. It used to be, back in the 1950s and 1960s, that more men died each year than women, usually by around 10 000 to 15 000 a year, or 200 to 300 a week. By the 1990s, however, the difference had narrowed, and then by the 2000s a few more women than men are dying most weeks.

We can work out the 2014-18 five-year average for each gender. For men for a seven-day period it’s 4951 deaths. For women it’s 5141 deaths. That means that on average 190 more women died per week than men in 2014-18.

But the five-year 2014-18 average is also somewhat misleading, because the gap has closed again in the last couple of years. If we look at weeks 1 to 16 for 2019 (the ONS gives us weekly figures for 2019, but not a sex breakdown for the whole of 2019), and add up the age breakdowns for each sex, we get 87 174 males deaths, or an average of 5448 per week, and 88 286 female deaths, an average of 5518 per week. That’s a difference of only 70 more female deaths per week.

Alternatively, we could do a three-year average for week 16 (2017-19), which gives us 4956.67 male deaths, and 5005 female deaths a week, a difference of 48.33 per week. Or, to extend it somewhat, a three-year average for weeks 15 to 17, which gives us 5095.44 male deaths, and 5153.55 female a week, a difference of approximately 58 per week.

Whichever of these we use as our baseline for week 16 it is apparent that we could normally expect between around 50 to 200 more female than male deaths in that week with the normal deaths. I shall assume a number near the middle of this range: We should expect 120 more female deaths in week 16, with the caveat that it would be normal for this to go up or down a few hundred either way.

So with this in mind let us return to our calculations. We can expect the 10 497 normal deaths for week 16 (the five-year average) to break down as follows: 5188.5 for men, and 5308.5 for women. When you add 5188.5 to the 1788 figure we got above (the expected excess non-Covid male deaths if they were all really Covid) you get 6976.5 for men, and when you add 5308.5 to 1308 (the expected excess non-Covid female deaths if they were all really Covid) you get 6616.5 for women. So these are the amounts we’d roughly expect to get for week 16 for non-Covid deaths if the excess deaths in that category were in fact undiagnosed Covid-19:

Male: 6976.5

Female: 6616.5

Difference: 360

What is the actual sex breakdown for non-Covid deaths for week 16? This is not released by the ONS. The ONS don’t even release the overall numbers of male and female all-cause deaths, or the overall male and female Covid-19 death numbers. But they do release sex breakdown by age numbers on their weekly deaths spreadsheet. If we add these up for all-cause deaths we get 11 445 male deaths, and 10 906 female. For Covid-19 we get 5057 male and 3701 female. Taking the latter numbers away from the former gives us the following numbers of actual non-Covid deaths for week 16:

Male: 6388

Female: 7205

Difference: 817 more female deaths

These numbers are significantly different to what we would expect from our workings out above (ie. 6976.5 men and 6616.5 women). The men are 588.5 less than we’d expect, and the women are 588.5 more. To put it another way, we’d expect the non-Covid category to have something like 360 extra male deaths if the excess non-Covid deaths were really undiagnosed Covid deaths, but in fact we have nothing like that. Instead we have 817 more female deaths. For that to happen there must have been around 1177 more female than male deaths in the normal non-Covid category, that is, in the normal deaths you normally get in week 16. But nothing remotely like that difference between male and female deaths ever occurs normally. The biggest differences are almost always around 200. So I think we can safely say that in week 16 of 2020 there was no sex difference of 1177 in the normal death category. That means that it is very likely that most of the 3096 excess non-Covid deaths are not undiagnosed Covid.

We should also note that this large gender difference in deaths is remarkable in itself. 6388 male and 7205 female non-Covid deaths in week 16 – 817 more female than male deaths – is an unprecedented imbalance, even in non-Covid times, and it requires an explanation.

Another way to see how extraordinary these numbers are, if you’re finding the foregoing hard to follow, is to just to look at the overall male and female all-cause numbers for the week 16. The ONS doesn’t provide these, but we can work them out from the age breakdowns: 11 445 male and 10 906 female. That’s 51.2% to 48.8%, or 51-49 rounded. Yet we know that generally Covid-19 deaths are more like 60% (or even more) to 40% (or even less). So it’s clear that there must be a lot of excess female deaths occurring in the non-Covid category to bring the overall numbers back to nearly 50:50.

My conclusion is that, as it is likely that most of the 3096 excess deaths in week 16 are not caused by undiagnosed Covid-19 deaths, it is most likely they are caused instead by the lockdown and the frightening messages surrounding Covid-19, which seems to be affecting women more than men.

Even if we suppose that only half the excess 3096 deaths were undiagnosed Covid, we are still in very unlikely territory as this would only decrease the expected differences between the sexes by about 300, and we would still require around 900 more female deaths than males in the normal deaths, which never happens. Besides, if only half the excess non-Covid deaths are undiagnosed Covid, then half aren’t, and that means 1548 deaths were caused by lockdown in just one week.

You may be wondering what the situation is with the data from weeks 14 and 15, when we also saw large increases in deaths. For week 14, there were 2607 excess non-Covid deaths. If these had all been undiagnosed Covid, then we would have expected 573 more male deaths amongst that number (1590 vs 1017), and we would expect that difference to reduce by around 120 to around 453 once we added in the regular non-Covid deaths to it. This is, in fact, roughly what happened: we ended up with 424 more male than female deaths in the overall non-Covid category (6668 vs 6244). So in week 14 there wasn’t support for the view that the lockdown itself was causing deaths. (It still may have been doing so, though, if such deaths were not skewed to females in that early lockdown week. But we have no evidence to support that.)

In week 15 there were 1783 excess non-Covid deaths, and that should have resulted in 410 more male deaths if these had all been undiagnosed Covid deaths (1096.5 vs 686.5), which we could expect to reduce by around 120 to around 290 in overall non-Covid deaths. In fact, the difference overall was only 45, which suggests that not all of the excess non-Covid deaths in week 15 were really undiagnosed Covid deaths, and that lockdown deaths were occurring. These numbers, however, are all consistent with normal variation, so we cannot really say anything with confidence about week 15.

It is suggestive, though, to look at how the proportion of female non-Covid deaths has increased over this period. In week 14 there were more male deaths in the overall non-Covid category, 51.6% to 48.4% suggesting there were a few undiagnosed Covid-19 deaths in this early stage. In week 15 female deaths were slightly greater, by 50.2% to 49.8%. Then, as we saw, in week 16 female deaths increased their proportion again, even more this time, with 53% female and 47% male. Although it is possible that this is just random variation, it would have to be a very large and very unusual random variation, so it is more likely female deaths are increasing for a specific reason, and that reason will not be Covid-19. It will be the lockdown itself.

I am grateful to Dr Paul Yowell, Associate Professor of Law and Fellow at Oriel College Oxford, who had similar thoughts to me, for discussion about this issue.

Appendix A: the weekly figures, weeks 14-16, 2020

The week 16 figures for England and Wales (Source: ONS):

All-causes deaths:

Total: 22 351

Male: 11 445 (51.2%)

Female: 10 906 (48.8%)

Difference: 539

Five-year average: 10 497

Excess over five-year average: 11 854

 

Covid-19 deaths:

Total: 8758

Male: 5057 (57.74%)

Female: 3701 (42.25%)

 

Non-Covid deaths:

Total: 13 593

Male: 6388 (47%)

Female: 7205 (53%)

Difference: 817 more female

 

Excess non-Covid (over five-year average):

Total: 3096 (13 593 – 10 497)

Expected proportion of these deaths if all are undiagnosed Covid:

Male: 57.74%: should be 1788

Female: 42.25%: should be 1308

Difference 480 more male deaths

– the actual difference once non-excess non-Covid deaths added in: 817 more female

– turnaround required: 1297

 

The week 15 figures for England and Wales (Source: ONS):

All-causes deaths:

Total: 18 516

Male 9948 (53.7%)

Female: 8568 (46.3%)

Difference: 1201

Five-year average: 10 520

Excess over five-year average: 7996 (18 516 – 10 520)

 

Covid-19 deaths:

Total: 6213

Male: 3819 (61.5%)

Female: 2394 (38.5%)

Difference: 1425 (23%)

 

Non-Covid-19 deaths:

Total: 12 303 (18 516 – 6213)

Male: 6129 (49.8%)

Female: 6174 (50.2%)

Difference: 45

 

Excess non-Covid deaths (above five-year all-causes average):

Total: 1783 (12 303 – 10 520)

Expected proportion of these deaths if all are undiagnosed Covid:

Male: 61.5%: gives us 1096.5

Female: 38.5%: gives us 686.5

Difference: 410 more male

– the actual difference once non-excess non-Covid deaths added in: 45 more male

 

The week 14 figures for England and Wales (Source: ONS):

All-causes deaths:

Total: 16 387

Male: 8794 (53.7%)

Female: 7593 (46.3%)

Difference: 1201

Five-year average: 10 305

Excess over five-year average: 6082 (16 387 – 10 305)

 

Covid-19 deaths:

Total: 3475

Male: 2126 (61%)

Female: 1349 (39%)

Difference: 777 (22%)

 

Non-Covid-19 deaths:

Total: 12 912 (16 387 – 3475)

Male: 6668 (51.6%)

Female: 6244 (48.4%)

Difference: 424

 

Excess non-Covid deaths (above five-year all-causes average):

Total: 2607 (12 912 – 10 305)

Expected proportion of these deaths if all are undiagnosed Covid:

Male: 51.6%: gives us 1590

Female: 48.4%: gives us 1017

Difference: 573 more male

– the actual difference once non-excess non-Covid deaths added in: 424 more male

 

Appendix B: the annual numbers for 2014-18

2018:

Male: 267,960

Female: 273,629

 

2017:

Male: 262,678

Female: 270,575

 

2016:

Male: 257,811

Female: 267,237

 

2015:

Male: 257,207

Female: 272,448

 

2014: 

Male: 245,142

Female: 256,282

 

2014-18:

Male: 267,960 + 262,678 + 257,811 + 257,207 + 245,142 = 1,290,798 / 5 = 258 159.6

258 159.6 / 365 = 707.28 x 7 = 4951

Female: 273,629 + 270,575 + 267,237 + 272,448 + 256,282 = 1 340 171 / 5 = 268 034

268 034 / 365 = 734.4 x 7 = 5140.8

– difference of 189.8

 

Appendix C: weekly deaths by sex for weeks 1 – 16, 2019

2019:

 

Week 1

M: 5194

F: 5761

 

Week 2:

M: 6172

F: 6437

 

Week 3:

M: 5891

F: 5969

 

Week 4:

M: 5903

F: 5837

 

Week 5:

M: 5604

F: 5693

 

Week 6:

M: 5787

F: 5873

 

Week 7:

M: 5860

F: 5964

 

Week 8:

M: 5643

F: 5652

 

Week 9:

M: 5463

F: 5581

 

Week 10:

M: 5543

F: 5355

 

Week 11:

M: 5230

F: 5337

 

Week 12:

M: 5203

F: 5199

 

Week 13:

M: 4916

F: 4951

 

Week 14:

M: 5049

F: 5077

 

Week 15:

M: 5189

F: 5102

 

Week 16:

M: 4527

F: 4498

 

2019 Year to week 16:

M: 87 174 / 16 = 5448

F: 88 286 / 16 = 5518

– difference of 70 per week

 

Appendix D: weekly deaths by sex weeks 13-17, 2017-19

Three-year week 15-17 averages:

 

2019:

 

Week 15:

M: 5189

F: 5102

 

Week 16:

M: 4527

F: 4498

 

Week 17:

M: 5021

F: 5038

 

2018:

 

Week 15:

M: 6106

F: 6195

 

Week 16:

M: 5566

F: 5657

 

Week 17:

M: 5064

F: 5242

 

2017:

 

Week 15:

M: 4161

F: 4330

 

Week 16:

M: 4777

F: 4860

 

Week 17:

M: 5448

F: 5460

 

Overall weeks 15-17 for 2017-19:

M: 45 859 / 9 = 5095.44

F: 46 382 / 9 = 5153.55

– difference of app. 58 per week

 

Appendix E: figures for theory that half of 3096 excess week 16 non-Covid deaths are undiagnosed Covid

If only half our 3096 excess non-Covid deaths for week 16 were undiagnosed Covid deaths (ie. 1548), then we would expect that to break down as follows:

 

First half: 1548 weighted by 57.74% for men and 42.25% for women

Male: 893

Female: 654

 

Second half: 1548, divide roughly 50:50, but give females a few more. How many? Roughly 120 excess female deaths on average per week for figures around 10 000 deaths, so for app. 3000 deaths make it 40 more female deaths:

Male: 754

Female: 794

 

Totals for this 3096 excess:

Male: 893 + 754 = 1647

Female: 654 + 794 = 1448

Difference: 199

 

Remaining 10 497 non-Covid (dividing roughly 50:50, but giving females 120 more deaths):

Male: 5188.5

Female: 5308.5

 

Total expected non-Covid if half excess non-Covid is undiagnosed Covid:

Male: 1647 + 5188.5 = 6835.5

Female: 1448 + 5308.5 = 6756.5

Difference: 79 more male

 

Actual non-Covid breakdown:

Male: 6388

Female: 7205

Difference: 817 more female

– requires 896 more female deaths in normal death category than males

 

Appendix F: A possible alternative explanation

One other possible explanation for the large number of excess female non-Covid deaths in week 16 is that the sex imbalance is caused by delays in submitting death certificates for men, perhaps because more male deaths are going to post-mortem, which is delaying the submission of their death certificate to the ONS. However, I have no evidence that this is happening, and as the ONS releases its figures eleven days after the week it is concerned with then we can expect most certificates to have come in. Besides, if that is happening in any significant way it would have happened with weeks 14 and 15 as well, so we should expect that week 16 would have included late male certificates from these weeks, which should have boosted the male deaths numbers in week 16. (Remember that ONS numbers refer to the certificates that come into its offices in the week in question, rather than the week when the death actually took place.)