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

expert reaction to ONS stats on deaths where COVID-19 or suspected or COVID-19 was mentioned anywhere on the death certificate, including in combination with other health conditions

The Office for National Statistics (ONS) is now releasing figure for the number of deaths in the UK where COVID-19 or suspected or COVID-19 was mentioned anywhere on the death certificate, including in combination with other health conditions.


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

“The ONS count deaths in which COVID-19 was registered on the death certificate – these can take 10 days to get through the system and be reported.  But it does mean that when they look back, they can say that about a quarter more deaths have occurred outside of hospital than have already been reported as deaths in hospitalised people (based on around 200 deaths).  This would suggest about 20% of deaths are happening outside hospital.

“So we might need to add roughly 25% to the number of deaths reported in hospitals to get the true number of deaths.  If that percentage stays the same over time it does mean the numbers reported in hospitalised people can be used for monitoring trends, even if it is an incomplete count of deaths.

“Over the last week the average daily increase in UK deaths has been about 21%.  What we are looking for is for that to slow down and flatten off, as is happening in Italy.

“It was running at a 30% daily increase in the UK – it seems to have declined a bit to around 20%, but there is a lot of uncertainty in that.  There are hints that the growth rate is slowing down – of course we want it to get to 0% but if we are following the pattern of Italy we can’t really expect that for another week or two.

“I don’t know what other countries report as deaths, and it’s not easy to work out what is happening in the UK.  The PHE ‘dashboard’ states that they are counting deaths with COVID-19 – if someone tested positive for the virus before they died it will be recorded as a COVID-19 death whether or not it was actually a cause of death.  But the deaths coming through ONS are based on Covid-19 being mentioned on the death certificate.  So they seem to be counting rather different things – some more clarity would be very welcome.

“Other countries might be more specific and only record a death if they know COVID-19 was a primary cause of death, rather than recording everyone who died with the virus.  This is important because we know most people who succumb to the virus have underlying conditions and are more likely to die.

“Whatever risk you have normally, the virus pumps up your risk – for a few weeks you are operating at about 15 times your normal risk of death.  While you’ve got the virus you are experiencing roughly  an entire year’s worth of risk in a short period  – that explains why most people get over it (generally those who would have likely survived the next year) and sadly others will succumb (generally if they were less likely to live out the next year).”


Comments sent out on Tuesday 31 March

Dr Adam Jacobs, Associate Director of Biostatistics at Premier Research, said:

“There are limitations in both sets of numbers used.

“The DH numbers only included patients who died in hospital after being tested for Covid-19 and having a positive result.  So patients who died at home or in a care home or a hospice would not be included in their statistics.  They would also not include patients who had all the symptoms of Covid-19, but, for whatever reason, hadn’t been tested for it (I don’t know how many patients come into that category).

“The ONS figures include all *registered* deaths where Covid-19 was mentioned on the death certificate, irrespective of where they occurred.  There are two limitations of this.  First, a doctor could potentially write Covid-19 on the death certificate in the absence of testing just based on the clinical symptoms.  Again, I don’t know how common that would be.  The other limitation is that deaths are not necessarily registered immediately.  I gather that the ONS have reported the number of deaths that *occurred* up to 20 March, based on deaths *registered* up to 25 March.  It is possible that many deaths also occurred before 20 March but had not been registered by 25 March, so are not counted in the ONS figures.

“According to this 2019 article, 39% of deaths are not registered within 5 days, so the ONS figures could be a considerable underestimate.

“So the two sets of figures are really not counting the deaths in the same way.  Both numbers are likely to be underestimates, but for different reasons.”


Prof Robert Dingwall, Professor of Sociology, Nottingham Trent University, said:

“Recording deaths is not a simple matter – and sociologists have been studying the process since the 1960s.  There are decisions about how to code the causes, checks and processing lags that mean short-term data can be quite problematic.  The UK system is not designed to produce accurate results at speed, unlike the French system, but to pick up unnatural causes for further investigation.  All recording systems have biases and limitations – it is important to know what they are.”


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

“I welcome today’s release by Office for National Statistics (ONS) which reports, by sex and age-group, the number of COVID-19-mention deaths in England & Wales that were registered in Week 12 (ends 20 March 2020: 103, with 108 registered in or before week 12, of whom one was younger than 45 years; 7 were aged 45-64 years; 21 aged 65-74 years; 34 aged 75-84 years; and 45 aged 85+ years).  Since around 18% of our population is aged 65+ years, the early toll in registered COVID-19-mention deaths had been around 12 times higher per million of population for the over 65s than for younger persons.  Safeguarding of the older population may change this in the weeks to come.

“ONS warns its readers that, due to some delay in death-registration being inevitable, the number of COVID-19-mention deaths that occurred by the end of week 12 but were registered after 20 March 2020 was 181 (and still counting).

“The next ONS release might usefully publish COVID-mention-deaths by death-week [i.e. in week 10, week 11, week 12] which had been registered respectively by the end of week 12, 13 and 14 [ends 3 April 2020], thereby affording like-with-like comparison because each week would have had the same registration-window; and to provide these counts by death-week, sex and age-group.

“Coroner-referred deaths that are not registered within 14 days of death having been ascertained may include some COVID-19-mention deaths.  An adhoc registration-route  would be needed for us to ascertain such deaths in England, Wales and Northern Ireland.  By contrast, fact-of-death has to registered in Scotland for all deaths within 8 days of death having been ascertained.”


Prof David Leon, Professor of Epidemiology, London School of Hygiene & Tropical Medicine, said:

“What ONS has done is important as it starts to provide a more complete picture of the impact of COVID-19 on mortality.  The value of these data will increase with time.  However, the caveats that ONS provide concerning inevitable (but relatively short) registration delays are important, and will explain the reason why counts are lower than the current government figures.  However, deaths which have COVID-19 as a cause on the death certificate that occur among people who were not tested for COVID-19 may misclassify them as such.  On the other hand untested deaths that were precipitated by COVID-19 may still go unrecognised.”


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

“I support the reporting of new figures but there is now a danger of the death figures becoming increasingly difficult to interpret.  We now have figures for the UK, England and Wales and the four separate administrations.  There are differences in the methodologies.  It would be useful if each separate reporting body produces a table or graph each time for their own figures as the figures cannot compared with each other.  Surveillance is about monitoring trends and this can only be done with consistent data sources.  An example of this is when China added CT findings to their figures and number of cases jumped overnight but the epidemic had not changed.”


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