select search filters
roundups & rapid reactions
before the headlines
Fiona fox's blog

expert reaction to updated PHE measures to record COVID-19 deaths

Public Health England (PHE) have updated they way that they record COVID-19 deaths.


Prof Paul Hunter, Professor in Medicine, UEA, said:

“The UK government has just changed the way that the daily reports of new deaths from COVID-19. Currently there are several reporting streams that produce data on deaths from COVID-19:

  1. The daily deaths for the UK which are published on the UK COVID-19 dashboard
  2. Daily reports on deaths in NHS Hospitals in England
  3. Weekly reports of deaths registered where COVID-19 was stated on the death certificate
  4. Excess deaths which the number of deaths that exceed the mean number of deaths for that week over the average for the past 5 years

“It is the first stream that has been altered, initially this included all deaths in anyone who had previously tested positive for COVID 19. Now this will only include deaths within 28 days after test. In addition England will also publish deaths occurring within 60 days or later if COVID 19 appears on the medical certificate of cause of death.

“None of these approaches can be said to give the actual number of deaths from COVID-19 and all give significantly different estimates.

“Currently the most timely report is probably those from NHS England, though of course there represent only a proportion of deaths as deaths have occurred outside of hospital especially in care homes.

“By using a 28 day cut off the new daily figures will not record those approximate 12% of patients who die from COVID-19 after this period. As for example a person who gets ill, is admitted to ITU and then dies some  5 or 6 weeks later.

“Whilst the ONS weekly death certification reports normally give the most definitive estimate of deaths, there is a delay between a death happening and that death being registered which means that the results when eventually published a two or more weeks out of date. Also during March and April there were many more excess deaths IN England and Wales than COVID 19 deaths reported by ONS. From early May when the availability of COVID-19 testing increased more COVID 19 deaths have been reported each week than the estimated excess number of deaths. Whilst it is not absolutely clear what proportion of the excess deaths in March April were from COVID-19 the fact that they fell below reported deaths at a time when testing became more easily available suggests that the majority of them represent undiagnosed COVID-19 deaths.

“If we compare the UK daily data with the ONS data for England and Wales up to 31st July there were 51,210 deaths registered where COVID-19 was listed on the medical certificate of death as a cause/contributing factor) Using the new daily reported figures for COVID deaths with the new data 41,189 were reported (across the whole of the UK) up to 31st July. This is a large discrepancy especially when considering that the ONS data is just for England and Wales and is probably also an underestimate.

“So the recent change has meant that the new figures on total deaths from COVID-19 which were always a fairly substantial underestimate of total deaths from COVID-19 are even more of an under-estimate. Nevertheless, they may give a more timely indication of whether or not deaths are increasing again than the ONS data.

“Having said this it is doubtful that we will ever get an accurate count on the number of people who did die from COVID-19. Probably the best estimate will come from some combination of excess deaths and death certificate reports. Looking at the ONS data this would give a figure somewhere between 60,000 and 65,000 deaths where the number of deaths in each week is the reported COVID 19 deaths or the estimated excess deaths, whichever is greater.”


Prof Sheila Bird, Formerly Programme Leader, MRC Biostatistics Unit, University of Cambridge, said:

“Contributing to the apparently slower rate of decline of COVID-19 deaths in England than in Scotland was that, in recent weeks, COVID-19 deaths which had occurred at 29-60 days after first positive swab-date were half the reported deaths.

“Public Health England was right to recognize, and count, COVID-19 deaths which occur more than 28 days after first positive swab-date but failed to account for them separately from COVID-19 deaths which occurred acutely. Yesterday’s report showed that, as of 3 August 2020, almost three-quarters (73%) of 1880 deaths (nearest 10) in England which occurred 29-42 days after first positive swab-date were COVID-mention deaths according to the Office for National Statistics (ONS), as were almost half (48%) of 1150 deaths during 43-60 days but only around 30% of 1200 deaths beyond 60 days.

“Almost all deaths occurring more than 60 days after first positive swab-test, and ONS-registered by 3 August 2020, will pertain to persons whose swab-test was prior to 1 June 2020.

“Let me offer to print and television journalists a suggestion for plotting the acute, mid-term and longer-term COVID deaths after first positive swab-test that I did not make explicitly (mea culpa) in my peer-review for Public Health England!

“Plot deaths in three coloured tiers against swab-date (or swab-week). The highest tier (Munros: yellow) counts the acute deaths which occur within 28 days of swab-date. The next tier (Corbetts: purple) are the mid-term deaths which occur within 29-60 days of swab-date. Both are already fully observed for persons whose first positive swab-date was prior to June 2020. The third and currently lowest tier (Grahams: amber) counts deaths which occur 61-365 days after first positive swab-date and are COVID-mention according to ONS. The third tier is accumulating but will be incompletely observed until 2021. For example, those whose first positive swab-date was 1 March have been followed-up for mortality for only 30+30+31+30+31+3 = 155 days as of 3 August 2020.

“We should, of course, anticipate that the age-distribution of those who die in the Corbett range may be different, younger for example, than those whose COVID-mention death occurs within 28 days. As the pandemic evolves (as indexed by later 1st positive swab-date), rehabilitation of patients may improve so that the ratio to deaths (acute to mid-term to longer-term; or Munros: Corbetts and Corbetts: Grahams) may also evolve as knowledge of how to manage the late sequelae of COVID-19 disease improves.

“A single cut-off is the wrong answer. Differentiating acute from mid-term and longer-term COVID deaths is the right solution.

“I now look to Public Health England and Public Health Scotland and to UK’s astute data-journalists and television-journalists to display these three tiers of COVID-deaths against first positive swab-date. Doing so will reveal that we have already crested our second wave of COVID-deaths (mid-term deaths) and are beginning to find our way in the Graham range. “Let’s put an end to naïve single number reporting in the face of immunological complexity and epidemic uncertainty.”


Prof Sir David Spiegelhalter, Chair of the Winton Centre for Risk and Evidence Communication, University of Cambridge, said:

“The 28-day limit marks an improvement over what is currently reported, but does exclude those who die more than a month after testing, even if they have Covid on the death certificate.  Including deaths up to 60 days, and later Covid-registered deaths, seems even better, but still excludes people who were not tested.

“This is a complex area and there is no truly ‘correct’ count.  The ONS figure based on registrations should still be considered the best available, but of course even this does not include additional excess deaths that do not have Covid on their certificate.

“My real concern is with communication to the media and the public. PHE have been consistently poor in clarifying both who is included as a Covid death and the inevitable delays in reporting, and their dashboard has given the strong impression that the daily count is the actual number of deaths the day before.  This in turn has influenced how the number is reported by the media, and is deeply misleading at this stage of the epidemic when the reported and the actual daily counts can be very different.

“I desperately hope that PHE can do more to prevent misinterpretation, but they have got themselves into a difficult situation in adding even more complexity to the multiplicity of Covid death statistics.

“My guess is that people would like an estimate of the actual number of deaths each day, even if this has to be revised. This is done by NHS England for Covid deaths in hospitals, and it is standard to revise statistics such as GDP as more information is obtained.  I hope that PHE might attempt such an estimate.”


Prof Carl Heneghan, Director of the Centre for Evidence-Based Medicine and Dr Jason Oke, Senior Statistician at the Nuffield Department of Primary Care Health Sciences, University of Oxford, said:

“Public Health England has changed its definition of deaths – here’s what it means:”


Prof Keith Neal, Emeritus Professor of the Epidemiology of Infectious Diseases, University of Nottingham, said:

“The two new measures are sensible.

“The 28 days is widely used in many countries and England is now the same as the rest of the UK. 

“One issue with 28 days is that some people are in intensive care for longer than 28 days and will not require a further test (or it may be negative later in the illness) and can clearly die from COVID.  The new 60 day limit requires COVID-19 to be on the death certificate so these COVID deaths are appropriate. 

“The previous measure of always being a COVID death even if recovered was unscientific.  As COVID deaths fall the number of recovered patients, particularly the very old and those with severe underlying conditions are now dying from these conditions and not COVID-19.  These non COVID deaths in survivors would become an ever increasing percentage of the England COVID deaths being reported.  It had become essentially useless for epidemiological monitoring.”



‘Technical summary- Public Health England data series on deaths in people with COVID-19’ (Updated 12 August 2020)


All our previous output on this subject can be seen at this weblink:


Declared Interests

None received.

in this section

filter RoundUps by year

search by tag