Deaths in London

A quick analysis of London mortality of spring 2020

In spring 2020, London was remarkably different from all the other regions of England and Wales in terms of all-cause mortality so it should be analysed on its own.

---

It is apparent that London suffered more excess death than could be explained by COVID.

---

However, when using a continuous distribution to model London’s excess deaths, calibrated between 15th and 29th March, it is apparent that there were ultimately fewer deaths than predicted.

The question is:

Is this down to the success of COVID intervention policies or something else?

---

Using the same model on COVID deaths, it would appear that there were in fact more COVID deaths than expected. Given that the first major intervention (school closures) did not occur until 19th March, we should not expect to see any impact until at least week ending 5th April given the observed time between infection and death so, whatever changes in activity occurred in London resulted in worse outcomes, not better.

The logical argument would be that if interventions had been made sooner and/or harder, outcomes would have been worse still.

---

Conversely, non-COVID excess deaths drop off sharply from expectations in the week ending 19th April (having begun their decline a couple of weeks earlier) and this is what is responsible for overall deaths ultimately being lower than first expected. This is also what we should see if a beneficial intervention is made (or indeed a harmful intervention is ceased).

As far as I can recall, there were no positive interventions that might reduce non-COVID deaths introduced in the week of 13th April. Therefore, some harmful activity must have ceased.

---

This is the sort of investigation that should be made in a public enquiry.

---

Updated Discussion

Thanks to Mike Yeadon for pointing out what really should have been more obvious to me.

Given the increase in COVID deaths over and above expectations at the same time that non-COVID excess drops sharply, the most plausible expansion is misattribution.

Both distributions start to deviate in opposite directions on 5th April and there is more than enough extra COVID death to account for the fewer non-COVID excess deaths.

The remaining additional COVID deaths would be offset by non-excess, non-COVID deaths.

So, the change in ”intervention” was simply the coding process, i.e. the over-zealous attribution of COVID in preference to any other cause.

In terms of public health, this alone can have seriously bad consequences as focus is shifted from diagnosing other, potentially treatable illnesses, much as we are now seeing in reverse with respect to the mRNA product.

Nevertheless, this does not fully account for the fewer than expected excess deaths overall. There was still another change between 15th March to 5th April and afterwards that resulted in fewer than expected deaths overall. All we can say with any certainty is that it wasn't a benefit of COVID measures.

---

I also considered the possibility that COVID was under attributed in the first few weeks but this does not fit as well with the rest of the data.

Finally, the data may well be incomplete. The ONS stop updating the dataset six months after year end so any deaths occurring after that date will be missing. These would actually be the most important deaths as registration is likely delayed due to referral to the coroner.

This issue is easily resolved with proper public access to the real time database instead of ad hoc data dumps. I have made my recommendation to the UK Statistics Authority.

---

Finally, given that deaths only start in London in the week ending 22-Mar-20, just before lockdown, I remain ambivalent about the likelihood that COVID deaths are genuinely due to the virus as opposed to disruption to the healthcare and other social support networks. There was some disruption before the official announcement on 22-Mar-20 that might push me into the latter hypothesis but it is not conclusive in my opinion.