Confusing and contradictory data is triggering pointless local lockdowns

A kind of data fever is gripping Britain, with numerous institutions pushing out daily surveillance reports, analyses, models and projections, in an attempt to get a handle on the pandemic.

If it all seems baffling and uncoordinated, then that’s because it is. Estimates between groups vary widely, and government policy appears to be driven by the flimsiest of results.

After a northern lockdown was imposed, chief medical officer Chris Whitty warned that Britain may have reached the limits of lifting restrictions, leading to fears that pubs and shops may have to shut so schools can reopen in September.

But what is the data actually showing and how reliable is it?

Take the weekly pilot infection survey’ results from the Office for National Statistics (ONS) that prompted Boris Johnson to bring in new restrictions in the north west at the beginning of August.

The ONS estimated that 35,700 people in England were infected with Covid-19 between July 20 and July 26, or 1 in 1,500 people.

The week before statisticians had calculated about 27,700 were infected, or about 1 in 2,000. Warning lights flashed and widespread hand-wringing commenced.

However, the new calculation was based on just 59 people testing positive out of a six weekly total of 116,026 swab tests. The previous week just 45 people tested positive out of a running total of 114,674. So just 14 extra positive tests had triggered mass panic.

Oddly enough, when the figures fell to 1,100 in June, journalists were told to ignore it as a data blip, and we have been regularly warned that determining infection rates on such low ONS numbers is fraught with difficulty. This time, however, the ONS said it "suggest" a "slight" increase.

“When ONS  stated ‘suggests’ it’s a signal something doesn’t add up,” said one Oxford academic.

Unsurprisingly, the hugely erratic figures swung back the other way on Friday, with the ONS estimating that 1 in 1,900 people are now infected, or 28,300 in total - 3,700 new cases a day.

Again, though, it was based on just 53 people testing positive, so just six people fewer than the previous week. Hardly a conclusive finding.

Meanwhile the King’s College app, which has been tracking symptoms from March and now has 3.5 million users, paints a much more optimistic picture. The data set is the largest out there and now includes swab testing. Yet its estimates have consistently been lower than most.

The app estimates there are about 26,500 infections in the whole of the UK, and 1,439 new cases per day in England - nearly three times lower than the ONS figures than prompted northern lockdown. King's researchers believe that cases have been stable for several weeks with no evidence of a second wave.

Nowcasting by Public Health England (PHE) and the University of Cambridge also suggests cases are flatlining, or possibly falling slightly.

But it is not just surveillance data that is muddying the waters. Trying to establish trends from the number of people testing positive for coronavirus is also bedevilled with problems.

The decision to lockdown in the north was also based on community and healthcare testing that appeared to be trending upwards.

Through July, testing showed a rise from about 500 cases a day to 750. However when picked apart, it became clear that Pillar 2 (tests in the community) were trending up while Pillar 1 (tests in healthcare settings) were trending down.

While more cases in the community might seem problematic, there was no equivalent rise in hospital admissions or death rates, suggesting it was mild cases that were being spotted. The number of very sick people continues to fall so the NHS is in no danger of being overwhelmed.

Look a bit more closely at the Pillar 1 rise and it becomes clear that the increase may actually be a result of more testing. Between July 1 and July 31, the seven-day moving average for Pillar 1 tests jumped by 20 per cent and Pillar 2 by 82 per cent.

When the rise in positive tests is mapped against the rise in testing, Pillar 2 is shown as flatlining and Pillar 1 declines a little more sharply.

Test, track and trace is also likely to be skewing the figures, particularly in hotspots like Leicester.

It stands to reason that testing people who have been in contact with those known to be infected will find more infections. That’s the whole point after all. It will pick up those who did not even know they have the disease and would never have been tested normally. It should be possible to account for this bias, but it is not being done.

It is also unclear how reliable the government’s figures are. On July 28, Pillar 1 reported 64 cases and Pillar 2 512 cases - 576 in total,  yet the government website reported a total of 547.

In May, it emerged that tens of thousands of cases had been double counted, and it appears that it is still a problem. When local lockdowns are being determined on low numbers to begin with, a missing 29 cases-a-day really matters.

Overall there seems to be no evidence of a second wave. Figures out from Imperial this week suggested cases had continued to fall after the easing of lockdown restrictions.

One in two hospital trusts have not reported a death in the past seven days and three quarters have not seen a death in the past 48 hours. Despite differences in daily rates, all groups seem to agree that about 8 people in 10,000 are infected and the R rate is unlikely to be above 1 for England.

Professor Karol Sikora, dean of medicine at the University of Buckingham, said: “At the peak there were around 3,000 coroanvirus patients admitted to hospital in England. The number is now around 60. There are 694 patients in hospital, down from 17,000. These are huge drops, and most importantly, they are continuing to fall.”