Fixes, Fiddles and Daily Figures
The Guardian reports there were “nearly 1,500 deaths in one day”. This figure is higher than the daily COVID-19 measure. Sir David King said it was “an attempt to play down the adversity that the country was faced with”.
The front page article led to claims of ‘fixing’ and ‘fiddling’ the daily COVID-19 figures. This article examines these claims.
Differences in measures
The Guardian looks at differences in COVID-19 death measures. The “actual death toll”, here, comes from the three statistical offices:
On 9 April the foreign secretary, Dominic Raab, standing in for the hospitalised prime minister, said the death toll had increased by 881 on the previous day. The actual death toll was 64% higher than that.
We can look at their definitions in detail:
- DHSC: Before 29th April, confirmed deaths with COVID-19 in hospitals in England. In Scotland and Northern Ireland, confirmed deaths in all settings. Public Health Wales requires clinical suspicion COVID-19 was causative factor. PHW covers confirmed deaths in hospitals or care homes. After 29th April, the England measure was: confirmed deaths in all places. The measure counts deaths by date of report, not date of death.
- ONS/NRS/NISRA: counts of death certificates where COVID-19 is mentioned. That mention does not need a positive test result. It only needs clinical suspicion. Statistics offices report these deaths by registration date and by date of occurrence. It takes time to certify deaths.
Both the DHSC and ONS measure associated deaths: deaths with the disease. This is different to deaths from COVID-19.
There are several key differences between the two measures:
- Date of report: The DHSC count is adding confirmed deaths in the last 24-hour reporting period. Those deaths were not in the last 24 hours.
- Coverage: Before 29th April, confirmed deaths in England referred only to hospital deaths. The data source was NHS England. Public Health Wales reports on deaths in hospitals and care homes.
- Suspected deaths: The ONS measure is for death certificate mentions. These deaths do not need to be lab-confirmed. A doctor can suspect the deceased person had COVID-19.
Consider 8th April — now recognised as the peak of COVID-19 deaths in the UK. With reporting lags, it is likely the reported number would be less the deaths by date of occurrence. The major difference was coverage: lab-confirmed deaths in England were only in hospital.
Was this limitation explained?
Yes. There was a joint statement between the DHSC and ONS, published on 31st March. The Guardian article omits this published statement.
Also, registrations can take a long time. What the article calls “actual death toll” is a provisional certificate count by date of death. The statistics offices may further revise those figures.
What did the daily press conferences show?
Sir David King is the former Chief Scientific Adviser. Prof King leads the ‘Independent SAGE’ group. He is the lone critic:
They didn’t say we have to add on all these other numbers which would have been a more honest thing to say. They were saying things were more rosy than they actually were.
The crux of this criticism rests on what the daily press conferences showed. The UK government has a page showing all the slides from the Coronavirus press conferences. This timeline highlights major changes:
- 30th March: The slide shows a global comparison of COVID-19 deaths, on a log scale. Day 0 is when the country reported over 50 total deaths. The slide notes reporting lags, but not differences in definitions. Worldometer is the source for other countries.
- 8th April: The graph comparing international deaths now uses a linear scale.
- 15th April: Johns Hopkins University is now the source for other countries. The slide also shows the ONS death certificate count for England.
- 16th April: England is replaced by ‘UK (all settings)’. This is the sum of death certificates from the ONS, NRS, and NISRA. The graph shows a UK-wide measure of confirmed hospital deaths.
- 21st April: A weekly slide compares hospital deaths and all registered deaths with COVID-19. The slide notes the median registration delay.
- 22nd April: Two new slides are in: daily COVID-19 hospital deaths, and a seven-day rolling sum by country. The latter is only shown on this day.
- 28th April: For the first time, the international comparison notes differences in definitions.
- 29th April: The updated DHSC measure is shown in a new slide. The graph has a seven-day rolling average. There is a new slide comparing reported deaths per million people, on a log scale.
- 30th April: The slides showing UK hospital-only deaths and the reported deaths per million people are both removed.
- 5th May: The weekly comparison slide is replaced by ONS certified deaths by place. These figures are for England and Wales.
- 6th May: The slide showing reported deaths per million people returns.
- 12th May: The slides have a new look. Slide 7 shows the DHSC measure with a seven-day rolling average. The slide shows new reported deaths for the last reporting day and the cumulative total. Slide 8 has weekly COVID-19 deaths from the ONS, including by place. The text explains these figures are: certified deaths with confirmed or suspected COVID-19. That slide is weekly. For other days, the DHSC measure slide mentions the respective difference. There is an annex for statistical notes. There are no slides showing global comparisons.
The daily press conference slides underwent a large number of changes. Slides after 16th April show UK death certificate counts. This is inconsistent with Prof King’s claim that ministers sought to “play down” the number of deaths.
This does not mean these press conferences did a good job of explaining the figures. As a different Sir David said, it was often “number theatre”.
Prof Sir David Spiegelhalter (Cambridge) highlights strengths and limitations could be better explained. What the numbers represent matters. This is particularly true for COVID-19 test statistics. The UK government figure conflates tests carried out with test posted out.
The DHSC and ONS measures represent different things.
Before 29th April, the DHSC measure excluded non-hospital deaths in England. The Department updated its measure. It counts deaths with a lab-confirmation of SARS-CoV-2. The figures are by reporting date, not when the death occurred.
The ONS measure is of death certificates which mention COVID-19. There does not need to a lab confirmation: there only needs to be clinical suspicion. Some doctors may not give correct diagnoses. It is important to look at all-cause mortality too.
Each measure has different strengths and limitations.