4,000 Deaths A Day 'Implausible' Says Professor David Livermore


David Livermore, Professor of Medical Microbiology at the University of East Anglia, has told Lockdown Sceptics he cannot see how Covid deaths could possibly reach 4,000 a day, the number Sir Patrick Vallance flagged up on Saturday.


In an email exchange with me and Jon Dobinson, the head of Recovery, he made a number of good points.


4,000/day is three times higher than the peak rate for Brazil, which has a population three times greater than the UK’s and a pretty laissez faire attitude to the spread of COVID-19


4,000/day is three times the peak for India, which relaxed a (failing) lockdown in the teeth of a rising infection rate back in May or June. Admittedly, India has a younger population and one with (likely) a lot more non-specific immunity, but it also has a population that’s 20 times larger than the UK’s.


SAGE, in the document leaked to the Spectator (estimating 85,000 deaths through the winter) is working on an infection fatality rate of 0.7%. So, 4,000/day translates to ~570,000 infections per day (4000/0.007) some three to four weeks earlier, or around four million per week. If you use the Stanford IFR of 0.25, it’d equate to 1,600,000 infections per day or 11.2 million per week Given that the SAGE document also (rightly in my view) opines that early reinfection is v. unlikely, such death rates, in the unlikely event that they were to occur, are not sustainable.


In the NHS data presented at the press conference, the health service was on course to exceed the currently available number of beds on November 23rd. Lockdown 2 starts on November 5th, therefore new infections acquired on or before November should continue to result in hospital admissions for a further ~14-16 days (six days for symptom onset and eight days for the tiny percentage of those who are infected who require hospital care to be admitted to hospital). So even if lockdown 2.0 reduces transmission, it still won’t have an impact for between 14-16 days, i.e. November 18th–20th. By that date, it should be obvious whether the NHS is three-to-five days away from being overwhelmed. If it does look that way, lockdown sceptics will look very silly; if it doesn’t…


According to the above graph, the peak in deaths will occur in the earliest days of December – say the 4th. We know that that the time between infection and death with COVID-19 is typically about 22-26 days (the 14-16 days to hospital admission, then another 8-10 to death). Therefore, people dying on December 4th would have become infected on November 8th-12th. If you accept SAGE’s IFR of 0.7%, that would mean there would need to be 570,000 infections per day by November 8th-12th. That’s quite a leap, given that the highest current estimate (Imperial) is 96,000 per day, doubling every nine days. Even on Imperial’s estimate, you would only get 200,000 infections per day by November 9th, or about one third of what is needed for 4,000 deaths per day by December 4th. Needless to say, estimates of current infections by the ONS and Kings College London are about half those of Imperial, with longer doubling times, meaning that they are even more impossible to reconcile with the Cambridge/PHE death plot.


The figure of 4,000 deaths per day becomes even more implausible if you assume an IFR of 0.25 (Ioannidis) rather than 0.7%. If you plug that assumption in, you need 1.6 million infections per day by November 9th to give 4,000 deaths per day by December 4th.


It might be argued that you get 4,000 deaths per day because, by December 4th, the NHS has been overwhelmed, leading to deaths of critically ill patients who would otherwise survive. But this has not been stated to be an assumption.


As you can see from the above graph, the blue line – based on PHE/Cambridge’s model – shows ~1,000 deaths a day at the end of October/beginning of November, whereas the average for the past seven days has been 260, with no single day above 361.


As Prof Carl Heneghan says, the PHE/Cambridge model’s estimate of daily deaths is about four times too high.