‘Death toll graphs were wrong’
Government forced to reissue key charts used to justify second lockdown after admitting projected fatalities were overstated
OFFICIAL projections that pushed the country into a second lockdown have been quietly revised to no longer suggest deaths could soon overtake those at the peak of the first wave, The Daily Telegraph has learnt.
Graphs shown at a Downing Street press conference on television last Saturday indicated England would see up to 1,500 deaths a day by early December, far beyond numbers seen in the first wave. But documents released by the Government show that the figures were in fact too high and have now been “amended, after an error was found”. The revised forecast reduces the upper limit to 1,000 deaths a day by Dec 8 – on a par with April’s peak.
Presenting the graphs, Sir Patrick Vallance, the Chief Scientific Adviser, said figures over a six-week period presented “a very grim picture” with “greater certainty” than long-term modelling could provide. But the Government Office for Science has altered two slides, reducing the range for deaths and hospital admissions.
While the presentation suggested daily hospital admissions could reach 9,000, the upper end has since been cut to 6,000 on the updated slides.
It comes days after it emerged that separate modelling, showing a worstcase scenario of 4,000 deaths a day by the end of December, was based on outdated figures. These have also been amended. They prompted Theresa May to question the Government’s use of statistics and ask if “figures are chosen to support the policy rather than the policy being based on the figures”.
Prof Carl Heneghan, from Oxford University, said last night the graphs presented at the weekend were “riddled with errors”, raising concern that a desire for lockdown had seen forecasts “systematically” exaggerated.
Greg Clark, the chairman of the Commons science and technology committee, said the belated admission of errors was “of great concern”, especially as one had been used as “the key projection” in the case for lockdown to prevent the NHS being overwhelmed.
Steve Baker MP, given advance sight of the projections ahead of the briefing last weekend, warned: “Government must accept public confidence rests on not over-egging the pudding.”
Yesterday, the UK statistics watchdog criticised the Government for a lack of transparency about the data driving its lockdown policies.
Last Saturday, when Boris Johnson announced lockdown, Sir Patrick presented a series of slides on the outlook for the pandemic, including the now disputed 4,000 deaths graph.
On Tuesday, Sir Patrick and Prof Chris Whitty, the Chief Medical Officer, were grilled by the Commons science and technology committee about the use of the modelling scenarios.
Sir Patrick said he “regretted” it if he did not make clear the scenarios were models rather than projections and were “not as reliable” as the six-week forecasts he had presented. He told MPS: “The right graphs to focus on are the six-week medium-term forward projections,” and described the two slides on hospital admissions and deaths as “the ones that are important”.
Amid angry exchanges, Prof Whitty said he had “never used anything beyond six weeks in anything I have ever said to any minister on this issue”.
Prof Heneghan claimed the data had been used “systematically” to drive the country into lockdown. He said: “It worries me that on matters this important, we are finding the data is absolutely riddled with errors.”
Revisions to data presented to 14 million people on prime-time television should not be “snuck out”, he said, and urged ministers to be more transparent.
“Public compliance is essential and in due course we will need people to take the vaccine. That requires people to trust the Government,” he added.
Yesterday, at a Downing Street briefing, Sir Simon Stevens, head of the NHS, said the service was dealing with the equivalent of 22 hospitals of Covid patients. He pointedly contrasted the NHS data with other charts presented.
He said: “Those are facts, those are not projections, forecasts, speculation. Those are the patients in the hospital today. And… we already know what is likely to happen, because today’s infec
tion is the intensive care order book for a fortnight’s time.”
A government spokesman said: “The main consensus projection remains unaltered. The data still clearly show… without intervention we are likely to breach the first wave of hospital admissions and deaths in a matter of weeks.”
Meanwhile, Rishi Sunak confirmed furlough would be extended to March. And Matt Hancock, the Health Secretary, said people would be allowed to travel abroad for assisted dying purposes during the second lockdown.
Dominic Raab, the Foreign Secretary, announced he was self-isolating after coming in contact with a positive case.
The data failures in this pandemic have been considerable – continually they have overestimated the numbers going to die, miscategorised Covid-19 deaths, exaggerated the impact the virus has had on hospitals and missed 16,000 cases because of an Excel blunder.
Perhaps even worse, these failings do not come to light because of Government transparency but rather when painstaking analysis, eagle eyes and sometimes even the simplest of questions expose major faults.
Recent events show that nothing has been learnt and point to a system that is failing. It wasn’t that difficult to spot that the slides used by the Government in their press conference last Saturday were out of date. And that the 4,000 death “worst case scenario” – was not only incorrect (on Nov 1 just over 200 deaths occurred when the model was predicting 1,000 deaths) but also weeks out of date; there had already been two subsequent available updates that had substantially revised down the estimated number of deaths.
At this point we were puzzled. How could the Government’s approach to data at this vital time be so shoddy? Rather than being thoughtful, analytical and cautious, the Government’s approach has been to frantically rush out worst-case scenarios irrespective of the need for accuracy. At the point when it mattered most, how could such basic errors arise?
The Office for Statistics Regulation is also concerned, warning yesterday that the use of data has not consistently been supported by transparent information provided in a timely manner. And as a result, there is the potential to confuse the public.
We are also concerned – the use of data is not just confusing, the errors are positively misleading.
On Saturday evening Sir Patrick Vallance, the Chief Scientific Adviser, stood alongside the Prime Minister, at the Downing Street press conference, and presented a series of slides.
We were informed that according to “medium-term projections” covering six weeks, the UK could see up to 1,500 deaths a day, and 9,000 hospital admissions by Dec 8.
And yet, after a clamour to see the modelling published – the day before MPS were to vote on it – it contained a startling addendum. An error meant the upper range of figures were too high by roughly 500 deaths and 3,000 admissions to hospital per day. The differences in the upper bounds of the “scenarios based on assumptions” are so stark they would probably affect what one might infer from the analyses.
Yet, the Government does not think so, judging by the caption of the revised slides: “Plots on slides 4 and 5 have been amended after an error was found in the interquartile ranges for SPI-M medium term projections. This does not affect the insights that can be taken from these analysis.” [sic].
This week has seen death estimates invalidated as statements on case numbers from the Chief Medical Officer to the science and technology select committee required “clarification to avoid any misinterpretation”.
To understand and interpret the current Covid-19 data is proving nigh on impossible. The publication and presentation of data is confusing – at times it is unsound.
The methods and data underlying the models used to generate “scenarios based on assumptions” are inscrutable and decisions taken on their basis are unaccountable. While the fog of a
‘The Government has been rushing out worst-case scenarios irrespective of the need for accuracy’
pandemic can justify or at least excuse some of these mistakes, nothing can justify the lack of transparency and humility.
Theresa May, the former PM, stated: “For many people, it looks as if the figures are chosen to support the policy rather than the policy being based on the figures.”
The growing number of errors seem to occur in only one direction – the worst case scenario – which underpins the point. Undeclared changes in projections based on undeclared assumptions clearly mislead.
Key decisions can only occur when the underlying data has been made available and checked independently, when methods and key assumptions underlying any models have been published. Only when these have been communicated clearly and coherently can effective and equitable policy decisions be made.