By Timon Wapenaar
A piece in the New Statesman by Sam Bowman which ostensibly aims to debunk the “eight biggest Covid-sceptic myths” presents us with a pot-pourri of Twitter-troll inspired overripe low hanging fruit. No serious lockdown critic has ever framed the argument in the way Bowman represents it. Has Sunetra Gupta ever said that “we are overreacting to a virus which 99.5% of people will survive”? I doubt it. A Google search for “Sunetra Gupta” and the exact phrase “we are overreacting” yields Bowman’s own article as the first hit, and only five other hits in total, none of which contained the damning quote.
Likewise, there is not one lockdown sceptic of stature who says that “we aren’t seeing excess deaths”, or that “we’re witnessing a ‘casedemic’ of false positives from doing too many tests”. While there are assertions made about both excess deaths and a ‘casedemic’, their nature is much, much more nuanced than Bowman would have us believe. Indeed, his phrasing of the ‘casedemic’ is absolutely absurd. How on earth could any opponent worth arguing with believe something as patently stupid as “more tests result in more false positives”? He obviously hasn’t read or listened to Heneghan, or Jefferson, or Yeadon, or McKernan. Or Levitt, or Gupta, or Bhattacharya, or Kulldorff, or Sikora.
It is mala praxis to choose the worst possible version of your opponent’s argument, and yet this is what Bowman does. The refutations he offers might be useful for dunking on your brother-in-law at a family dinner, or vanquishing that Twitter troll with a hundred followers, but fall hopelessly short of being a contribution to any real argument over the efficacy of lockdowns.
If he’d been more courageous, Bowman would have addressed the most important question posed by sceptics: are state-mandated NPIs (non-pharmaceutical interventions) sufficiently effective to justify the significant harms they entail? That they reduce transmission to some degree (which they do, in all likelihood) is not sufficient. The gains must outweigh the price paid by society. Bowman instead chooses to focus only on whether lockdowns reduce transmission. This is disappointing: if the holistic effectiveness of the most restrictive lockdown measures can be proven, all other arguments become somewhat ancillary.
As evidence for the effectiveness of lockdowns in reducing transmission, Bowman offers two assertions:
1. Lockdown worked because data shows that cases began to decline at the same time or soon after lockdown was introduced.
2. Two studies produced modelling which demonstrated that lockdowns reduced infections.
Let’s start with the second contention, as the studies Bowman cites can actually be used to argue against the first.
More lockdown is not necessarily better lockdown. Brauner et al. and Haug et al. (the two studies cited by Bowman) each model the effects of state-mandated NPIs, and each make the point that effective outcomes can be achieved with a mix of measures significantly less intrusive than those which have been pursued to date. Both studies found that stay-at-home orders had a minimal effect on reducing R. Brauner, et al. goes so far as to claim that introducing just three NPIs (closing schools and universities, closing high risk businesses, and limiting gathering sizes to a maximum of 10) could be sufficient to reduce Rt to below 1 from the global median R0 of 3,3 (Brauner gives an R0 in the UK of 2.97). Imagine how different that would have been.
Nevertheless, these studies are far from conclusive, the authors repeatedly pointing out that geography, demographics, and specific implementation details have unmeasured effects. They also take pains to point out that their models do not account for changes in behaviour which might occur independently of NPIs. In the model created by Brauner et al., inputting an R0 of 3 and then removing all NPIs spits out an Rt of 3. In other words, the model assumes that a population threatened by a new pathogen would do nothing to protect itself unless the government forced it to do so.
But people do change their behaviour in the midst of a pandemic, and the data from the UK government’s coronavirus dashboard cited by Bowman proves it.
I fear Mr Bowman has forgotten his epidemiological priors: the overwhelming majority of people are not PCRed on the very day they become infected by COVID-19. Carl Heneghan, of Oxford’s Centre for Evidence Based Medicine points out the importance of Farr’s Law: there is a lag between infections, cases, and deaths . Brauner et al. assume a lag of 10.92 days from infection to case confirmation. With this in mind, let’s look at recent events in the UK.
Starting in early December, daily confirmed case rates began to rise in the UK. This rise continued, and on January 5th another full lockdown was declared. Case rates peaked on January 1st (if determined using a seven-day average), after which they began to decline again.
Bowman asserts that the former (reinstatement of full lockdown), led to the latter (a decline in cases), yet applying the 10-day lag between infection and cases we see that the rate of infection probably began to decline around December 22nd. Either Farr’s law doesn’t hold, and Brauner’s model (and virtually all other similar models) is broken due to a very flawed assumption, or the implementation of a full lockdown on January 5th had little to no effect. What effect it did have would have been seen in the data from January 10th onwards, at which point the fall in seven-day average case rates actually begins to flatten very slightly.
So what did happen to cause infections to roll over? One possible explanation is that Britons, anxious at rising case numbers, heeded the call to “Save Christmas” and began to change their behaviour before the Government forced them to do so.
Another possible explanation is that the wave peaked of its own accord – as all epidemic waves eventually do. Perhaps the decline was due to a combination of the two. Or due to some other, unknown factor. But obviously something which occurred before the thing it was supposed to have influenced.
We can apply the same logic to the UK’s first lockdown, as Carl Heneghan did. At that time, deaths were used to measure peak infections, as COVID testing was in its infancy and generally only done when people sought treatment. The lag from infection to death given in Brauner et al. is 22 days, and COVID deaths in the UK peaked on April 8th. This strongly suggests that infections peaked on or around March 17th, roughly ten days before lockdown was implemented.
So we can see that lockdown is very far from being a panacea. To achieve its stated purpose, it has to be wielded with the precision of a surgical tool, and the size and complexity of the undertaking makes that almost impossible. Real lockdown sceptics don’t address their criticism to the general public. Our quarrel is with those who would seek to wield that surgical tool in a clumsy and harmful way.
Thus, it was with a degree of bitterness that I read Bowman’s paragraph on age stratification in COVID-19 mortality. We all recall how health mandarins from London, to Madrid, to Bergamo, to New York, to Stockholm (yes, even in Stockholm) and beyond made the decision to deny hospitalisation to the elderly and to oblige care homes – which quickly became infectious foci – to shoulder the burden. They did this because it was part of their strategy for wielding that surgical tool (the aim was ostensibly to shield the hospitals), and the results speak for themselves: between March 2nd and June 12th, 29% of all deaths in care homes in the UK involved COVID-19. In total there were 30,000 excess deaths in care homes. If Bowman were intellectually honest, he would remove at least some of those deaths from his IFR calculations.
But the age stratification question goes even deeper. The median age of those in the UK who died from COVID-19 up to October 2nd was 83. Life expectancy in the UK is closer to 80. It turns out that in long-lived societies, the chief risk factor for mortality is advanced age, as it should be.
But what of the excess? Surely this points to something happening. And yes, it does. Once again, no lockdown sceptic worth his or her intellectual salt denies that there has been excess mortality, and that this excess mortality is largely being driven by deaths which have been recorded as COVID-19 deaths. But anyone vaguely familiar with the concept knows that excess mortality is almost perennial. What is at issue is the baseline, and the proportion of excess deaths which are attributable to COVID-19, versus those deaths which are attributable to the shocks imposed by lockdown. An excellent illustration of the problem can be seen in data from Sweden, where total excess mortality in a country which avoided implementing the strictest lockdown measures is substantially less than the total COVID-19 death toll.
Does anyone really suppose that the implementation of measures disruptive to society on the scale we are witnessing would not produce excess mortality in and of itself? The ONS data which Bowman shares on excess mortality relies on PCR testing to distinguish between COVID-19 deaths and non-COVID-19 deaths. Unfortunately, as we all know, there are significant problems with the PCR (Ct-hacking, problems with the Drosten assay, and Bayesian effects which occur when prevalence is low), none of which is addressed in the New Statesman piece.
In all likelihood, we will never know exactly how much excess mortality was due to COVID-19 pathology, and how much was due to the iatrogenic effects directly and exclusively owing to lockdown. Even the total number of excess deaths is very sensitive to the choice of baseline. A reasonable guess seems to be that as much as 30% of total all-cause excess mortality in 2020 was caused by lockdown. This can be inferred from location of death data as well as age stratification and conventional death reporting. But it’s still a guess.
We are on more solid ground when we look at nosocomial infections – those produced in hospitals – and we see that they are non-negligible, accounting for up to 25% of all infections in some areas of the UK. Once again, measures taken with the objective of wielding our surgical tool have strained hospitals, as increasing numbers of healthcare workers are sent home after testing positive, and patients who are admitted for something completely unrelated to COVID-19 must be moved to Covid wards if they become infected after admission.
So, we see that the complexity is such that significant feedback is produced, and measures designed to suppress transmission may also provoke it. Bowman’s (and the New Statesman’s) response to this is to reduce the two sides of the argument to a collection of Tweet-sized memes, and then, like the child with two insects trapped in a jar, to have them face off against each other. To be sure, this tactic is also used by the less nuanced lockdown sceptics, who are just as convinced that their low-resolution arguments are sufficient. Meanwhile, in the real world, the adults are having a quite different conversation.