The Long Tail

Deaths with COVID-19 show large differences by source.

Public Health England are reviewing their definition of COVID-19 deaths. This article compares their measure to figures from the Office for National Statistics.

Counting certificates

The Office for National Statistics publishes a weekly report on death registrations. These reports are for England and Wales. It takes time to register a death, and process that registration. As a result, the published figures covers registrations 11 days or more in the past.

What do Medical Certificates of Cause of Death show?

Clinicians must certify causes of death “to the best of their knowledge and belief”. Doctors are not infallible, nor expected to be.

The Medical Certificate has two parts, following World Health Organisation recommendations:

  • Part I: The sequence of diseases or conditions that led to the death.
  • Part II: Other significant factors, which contributed to the death.

The guidance gives an example:

Diabetes contributed to, but did not cause, this death. (Image: GOV.UK)

Colon cancer — listed in I(c) — caused liver metastases. Those deposits ruptured — in I(b) — and caused a fatal haemorrhage, I(a).

As the start of the causal chain, colon cancer was the underlying cause of death. Diabetes (II) contributed to the death.

The ONS has its measures of COVID-19 deaths:

  • Involving COVID-19: the certificate mentions COVID-19. This mention is either as a causal link or a contributory factor.
  • Due to COVID-19: the disease is the underlying cause of death.

These mentions do not need a lab-confirmation of SARS-CoV-2. Doctors may suspect someone has COVID-19 without a positive test result.

Also, a deceased person having a positive test may not feature on the death certificate at all. Clinicians may consider that diagnosis unrelated— neither causing nor contributing to the death.

Comparing measures

The daily measure of COVID-19 deaths in England differs. The Public Health England definition is:

  • Lab-confirmation: a positive test result for SARS-CoV-2.
  • All deaths: any death with an associated positive test result. That death can be in any setting in England, from any cause.

How big is the difference between the PHE and ONS counts?

The Office for National Statistics processed 48,528 certificates involving COVID-19 in England. These deaths occurred after 6th March and before 10th July. The registrations were by 18th July. (The ONS registered four deaths involving COVID-19 occurring before 5th March.)

For recorded deaths by 20th July, the Public Health England number was 40,382. That is a difference of over 8,000 deaths.

In the early part of the epidemic, testing volumes were low. Before 30th April, fewer than 50,000 tests were made available each day.

Some clinical diagnoses did not have lab-confirmation. Around the peak in deaths, this is the most plausible explanation of the difference.

I have written annotations to help the reader. (Image: R Pubs)

We compare ONS certificates registered up to 18th July, with PHE deaths recorded up to 20th July. Every day since 29th May: PHE deaths by date of occurrence exceed ONS certificate counts.

In effect, Public Health England creates a growing cohort of people to study. These are people in England who had a positive test result for SARS-CoV-2.

Their method checks if those people are alive or not. When those people die, that is a confirmed ‘COVID-19 death’. That way of counting creates a long tail.

Death certificates mention believed causal or contributory factors. A positive test result for SARS-CoV-2 may not be relevant to the death.

What about Wales?

Public Health Wales has a different definition for its daily measure:

  • Lab-confirmation: a positive test result for SARS-CoV-2;
  • Settings: in Welsh hospitals or care homes;
  • Cause: clinical suspicion COVID-19 was a causative factor in the death.

How big is the difference in Wales?

The ONS certificate count was 2,489. These deaths occurred up to 10th July. The registrations were by 18th July.

For records up to 20th July, the Public Health Wales death count was 1,542.

The difference is around 900, resulting from three aspects:

  • Suspected deaths: the ONS includes deaths where doctors suspect the deceased had COVID-19. This suspicion does not need lab-confirmation.
  • Place of death: the PHW count is only for deaths in hospitals or care homes. People may die in other places, such as at home.
  • Contributory factors: a death certificate may mention COVID-19 as a contributory factor. It does not need to cause the death.
It is more rare for PHW counts to exceed the ONS figures. (Image: R Pubs)

There are some days when the PHW measure exceeds the ONS certificate count. Since 29th May, there have been seven such days. This is not persistent, as it is in England.

It may arise due to clinical disagreements. A doctor can suspect COVID-19 may have caused the death. A different clinical opinion appears on the medical certificate. It may also be due to late registrations.

While less timely, statistical offices count and analyse death certificates. Those certificates feature clinical judgements about causes and contributors to how people died. For understanding epidemics, that analysis is crucial.

The R Code for the two graphs is available on GitHub and R Pubs. My previous article looked at differing definitions for daily measures of COVID-19 deaths.

This blog looks at the use of statistics in Britain and beyond. It is written by RSS Statistical Ambassador and Chartered Statistician @anthonybmasters.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store