No, The 'Flaw' In England's Covid Data Does Not Mean The Death Toll Is Overestimated

But the anomaly does need correcting because it does mean we are probably over-counting *recent* deaths – something that will only get worse.
|

Get the latest on coronavirus. Sign up to the Daily Brief for news, explainers, how-tos, opinion and more.

Public Health England (PHE) has come under fire for a “statistical flaw” in its coronavirus death figures meaning people who tested positive for the virus but later died due to unrelated causes are being identified as Covid-19 deaths. 

Health secretary Matt Hancock has ordered an urgent review into the way PHE counts fatalities, amid concerns that the statistics are exaggerating the true picture of coronavirus deaths in England. 

But this doesn’t mean the number of deaths in England since the start of the pandemic has been inflated, for the simple reason that the anomaly is dwarfed by the undercounting in other areas.

Here’s an explanation of what has happened with Public Health England’s data, and what it tells us about our understanding of the devastating impact of Covid-19: 

What has happened? 

Researchers have criticised “statistical flaws” in the way the deaths are reported across England, saying they are left looking far worse than any other part of the UK.

PHE’s figures feed into the daily death statistics published by the Department for Health and Social Care (DHSC). Data from Public Health Wales, Health Protection Scotland and the Northern Ireland Public Health Agency are also fed in.

The daily DHSC data represent the number of newly reported deaths of people who have previously tested positive for Covid-19, who have died in all settings.

But in a blog entitled “Why no one can ever recover from Covid-19 in England – a statistical anomaly”, Professors Yoon Loke, from the University of East Anglia, and Carl Heneghan, from the University of Oxford, said more robust data is needed.

They argued that PHE looks at whether a person has ever tested positive and whether they are still alive at a later date.

This means anyone who has ever tested positive for Covid-19 and then dies is included in the death figures, even if they have died from something else. They added that, for example, if someone recovered from Covid-19 but died after being hit by a bus months later, their death would still be recorded as a coronavirus death. 

Loke told HuffPost UK: “The problem is that when we look at the hospitals’ data we see that actually fewer and fewer people are dying in hospital due to coronavirus, and yet several days a week we see in the press that hundreds of people are still reported as having died outside of hospital. 

“A lot of members of the public are quite terrified by this and wondering: ‘Who are these people, and why are they dying?’

“When we looked at the data we realised that the NHS had actually been very successful and discharged about 80,000 patients back to their home but despite these patients having recovered PHE are still monitoring these people and if they die of an unrelated condition PHE will still put them in the daily figures.”

England, Wales, Scotland and Northern Ireland all report their deaths separately, but the three devolved nations have avoided PHE’s situation by implementing a 28-day cut off. No such cut-off point exists in England, meaning that – while medical records are maintained regarding discharge from hospital – from a statistical standpoint, nobody who tested positive for the virus ever recovers.

Loke said: “Many of the people who have had coronavirus and have been discharged are quite old and many of them will gradually die because of unrelated causes. So the number of coronavirus deaths in England will never get to zero – there will be a constant stream of deaths all the time based on the 80,000 or so hospitalised patients who have survived.

“The ONS have actually shown in the past few weeks that the rate of excess deaths is no different to the five-year average, so PHE are actually picking up part of the average death rate and putting that into the coronavirus figures. 

“That’s pretty scary for the public to be facing every day, especially when they look at Scotland, Northern Ireland and Wales and say: ‘Why? Why are these other places doing better?’ Actually, I don’t think they’re doing better – it’s just that Public Health England are recording them in quite a different way.” 

Loke and Heneghan concluded their article by noting that: “The Covid death toll in Britain up to July 2020 will eventually exceed 290k, if the follow-up of every test-positive patient is of long enough duration.”

On the government death statistics website for England, the issue is acknowledged, saying: “Deaths are counted where a lab-confirmed positive coronavirus test result is reported in any setting.

“This means that not all deaths reported here are caused by coronavirus.”

Open Image Modal
A statistical anomaly in the Public Health England coronavirus deaths data was uncovered on Friday.
STEVE PARSONS via Getty Images

Does it mean that the UK has far fewer deaths Covid-19 deaths than we thought? 

In the wake of the news of the statistical anomaly, the reaction of some was to claim that coronavirus deaths – a statistic that has formed so much of our perception of the virus and been used to justify lockdown measures – were far lower than the government had told us.

But the DHSC figures are not the only measure we have to understand the enormous impact of Covid-19 on the number of deaths in England and Wales. 

The most recent excess figures data published by the Office of National Statistics (ONS) show that there have been 53,979 “excess deaths” in England and Wales in 2020 so far – that is, the number of people who have died this year over and above the average for the last five years.

Of the deaths registered by July 3, 50,548 of them mentioned Covid-19 on the death certificate, accounting for almost 15% of all deaths registered in 2020. 

Yet the overall number of coronavirus deaths recorded by the Department of Health and Social Care (DHSC), which incorporates the PHE statistics that we now know to be affected by this statistical anomaly, currently stands at 45,119 for the entire UK – more than 8,000 lower than the ONS figures that cover just England and Wales. 

Of course, the PHE figures account for only England, while the ONS ones cover both England and Wales. According to Public Health Wales there have been 1,545 Covid-19 deaths in the country recorded since the start of the pandemic. Like Scotland and Northern Ireland, Wales also uses a 28-day cut-off period. 

It means that while the PHE figures are obviously problematic – and as the year continues they will become more unreliable due to the increased likelihood of someone dying of an unrelated cause after testing positive for Covid-19 – they don’t mean that we’ve actually had a lower number of deaths than published by the government.

The ONS Covid-19 mortality data is based on registrations of deaths where confirmed or suspected Covid-19 was mentioned on the death certificate, wherever the death took place.

The death certification used by ONS to measure Covid-19 deaths does not depend on a positive test, while the Public Health England figures are reliant upon this positive test – which is why they are so relatively low.

Due to the well-documented issues with accessing comprehensive testing, particularly at the height of the outbreak and in a care home setting, it has been widely accepted that not every person who died with Covid-19 had actually tested positive for the virus.

Nick Stripe, head of the ONS health analysis and life events division, told HuffPost UK: “This anomaly would help explain why PHE figures have been slightly higher than ours recently, when I’d expect them to be the same or slightly lower than ours still because of the dependency on a positive test result.

“Our statistics are the more complete measure. PHE stats are only where a death is linked to positive test. If the critique of PHE’s method is right, and it seems it might be, then my guess is it might mean a few hundred to a thousand of the deaths PHE have reported in recent weeks might need looking at and checking.

“It might not be a black and white issue if earlier Covid infections have resulted in longer-term issues which have resulted in death. But it would be a problem for those where the death is unrelated to an earlier positive Covid test.”

Loke added: “Early on we were not able to accurately capture all the coronavirus deaths – we were underestimating for probably the first two or three months because we did not have testing. 

“But now we’ve reached a point where there is so much testing available that we’ve actually swung to the other point now where if we use PHE data we’re actually overestimating deaths that have occurred over the past six weeks or so.

“If we’re looking at the total number of deaths then it is probably somewhere between what PHE and ONS are saying. The difficulty is with the timing of the deaths – we’ve swung from an underestimate to an overestimate. 

“Either way it’s problematic for the scientists, the clinicians and the public when they are interpreting the data.”

Do we know how many people have actually recovered from Covid-19? 

The issues uncovered in Friday’s CEBM article are further complicated by the way the UK tracks recovery rates – or the fact that they may not be monitored at all. 

To put it plainly, we don’t know how many people have had Covid-19 and recovered. Unlike many other countries, the UK government doesn’t publish data on recoveries and it is unclear how or if they are comprehensively tracked. 

When HuffPost UK contacted the DHSC in May to enquire about why the recovery rates had not been published, a department spokesperson said: “While some countries do publish this type of information it varies widely in what it actually shows. We are looking at a way based on the length of time it takes to recover.”

They added that different nations’ approaches to recovery figures “vary widely in methodology and approach”, which could lead to “very misleading” international comparisons. 

No information was provided on the specific nature of the new measure the government was developing to track recovery, and no date was given on when this data would be made public. It has yet to materialise two months on.