This is an article by Simon Anthony, a one-time theoretical physicist.
Hector adds: I don’t agree with all of Simon’s analysis here, but you don’t have to write something I agree with to get published here. If you’re thinking of submitting a piece, see the details here.
There’s been some disagreement about the justification for and consequences of lockdowns (he said, with a degree of understatement). In days, months and years to come, there’ll be furious arguments over whether lockdowns were necessary rather than less authoritarian measures. To begin to make the case for lockdowns, those in favour have, as an absolute minimum, to show that they worked in halting the spread of Covid-19.
The argument that lockdown will work
Fairly obviously, the purpose of a lockdown is to reduce the spread of the virus and consequent deaths by reducing the links between people. To reduce an R0 of 2.4 below 1 (the critical value below which the outbreak will be suppressed) to, say, 0.8, the average number of contacts to whom the virus is transferred has to be reduced by two thirds. The Imperial College paper which led to the UK’s lockdown mentions that, in their model, ‘approximately one third of transmission occurs in the household, one third in schools and workplaces and the remaining third in the community. These contact patterns reproduce those reported in social mixing surveys.’ So, if schools and workplaces are closed and community contacts are severed, that removes two thirds of contacts, leaving only the third of transmission which takes place in households. Essentially this is what the UK lockdown did.
The predicted effects of lockdown in the UK
For critics of the IC model, it’s worth examining in some detail the paper’s predictions. First, it predicted that if the measures of case isolation, home quarantine, social distancing for the elderly and school/university closure were combined, there would still have been ~250,000-350,000 deaths and ICU facilities would still not have been able to cope with peak demand. However, the paper went on to say that, if social distancing was extended to the entire population, not just the elderly, and the other restrictions kept in place, then deaths could be reduced to ~39,000 and ICU facilities may be sufficient. Such a level of social distancing would in practice require the closure or suspension of workplaces, social gatherings, restaurants, cinemas, sporting events etc, just as has been imposed under the current lockdown.
So IC’s prediction was that the lockdown would reduce the unmitigated hundreds of thousands of deaths to tens of thousands. (It’s slightly more complicated in that IC envisaged that lockdowns would be intermittent – triggered at particular levels of ICU admissions and then turned off when that number had dropped – so the figure of 39,000 deaths was cumulative over 2 years but with a large initial peak. This was under the least vigilant threshold which was however already exceeded when lockdown began in the UK. It’s therefore likely that many of the deaths which under Ferguson’s model happened later, were brought forward to the first peak which was consequently more lethal than the model envisaged. So I’ve compared Ferguson’s total over two years with the current UK peak) and that the then current ICU facilities would be – just – sufficient. Currently ~36,000 deaths in the UK have been attributed to Covid-19 and, although there was much talk a few weeks ago of ICUs being under pressure, it seems to have got by and not needed to use much, if any, of the emergency facilities which were hastily added.
Ferguson couldn’t be blamed for wishing that, rather than vehemently criticise the media’s version of his paper, people might actually read it and find that his predictions for the effects of lockdown really aren’t bad at all. In fact, given all the uncertainties, confusions and complications, so far at least they’ve been quite remarkably accurate.
In most other countries essentially the same lockdown pattern was followed, by a mixture of direct and indirect interventions, varying to a limited extent (in Sweden and Japan, for example) when local demographics (lower household sizes and population density in Sweden) or voluntary actions (again, Swedes seem to have social distanced readily; Japanese likewise) and cultural factors (Japan again) allowed the possibility. The outstanding exception seems to have been Taiwan which had previously developed systems and technology to prevent novel viral infections becoming established.
The evidence that lockdowns have worked
Overall, it’s not at all clear that there was really a great deal of variation between different countries’ lockdown protocols. Similar places closed and people socially distanced, whether by choice or imposition. If they worked as planned, such similar lockdowns should have had similar effects.
The chart above shows the deaths per day (averaged over the previous seven days so as to reduce weekend effects) for eight countries, starting from the day on which they imposed lockdown. The numbers on the left are the percentage of the maximum number of daily deaths for each country. (Sweden didn’t have a formal lockdown. Its start date is set as April 1 because on that day its government asked people to socially distance.)
It’s well known how varied the timing and allocation of reported Covid-19 deaths has been (hence the jagged appearance of the lines in the chart). The similarity of the shape of the lines for different countries despite such variation therefore seems to be persuasive evidence that the lockdowns have worked in much the same way wherever they’ve been applied. The overall pattern is common to all eight countries: a rise in the rate to a peak about ~3 weeks after lockdown, a few days close to the peak, then a decrease, at a lower rate than the initial rise.
Some people will be aware that, for example, the UK’s peak death rate is thought to have been April 8, sixteen days after lockdown. In the chart above, the peak is twenty-two days after lockdown. This discrepancy, and similar examples in other countries, is partly because I’ve used figures for daily announcements of deaths rather than the days on which they actually occurred and partly because the chart averages data over the previous week.
Proponents of lockdown will explain the shape of the curve as follows: before lockdown, Covid-19 was spreading exponentially. Over a period of a few days, lockdown halted this growth. Of those already infected, a proportion went on to die. The number dying each day grew exponentially for a time, just as the infections had done. On average, people who die from Covid-19 do so about 14 days after first showing symptoms. The average symptomless incubation period before then is ~5 days (these numbers are still debated). So the peak number of deaths per day should follow roughly 19 days after the peak of infections. As lockdowns took a day or more to fully come into effect, a lag of ~3 weeks would be expected between the peak of daily infections and that of deaths.
But of course these estimates can only be approximate. The date of lockdown may be misleadingly precise as it generally took place in stages and was preceded by a period of heightened anxiety and vigilance during which voluntary social distancing began. The obvious method of assessing the data of the peak of new infections – by counting the number of new cases – is very uncertain as it depends on the rate of testing, the conditions to be met for a test to be administered, variable delays between tests and reports… all of which differed between and within countries and which were constantly changing.
Comparison of R0
While the shape of the curves is similar, it would be useful to also compare the underlying numbers to see how well R0 matches (or differs) between the different countries. There are two distinct sections – the period up to the peak during which the mortality rate increased and R0 was greater than 1 and the period after the peak, when it was less than 1. For the reasons outlined above, the precise location of the peaks is imprecise so to calculate R0 for the upward slope I included points up to a week before the nominal peak and for the downward slope I included points starting a week after the peak. The table below shows the results:
There are several striking features among these numbers which are worth commenting on:
1. The first 7 countries have estimated R0s during the growth phase of between 2.0 and 2.8. An unmitigated outbreak with R0 greater than 2 would infect ~80% of the population.
2. The UK’s R0 during the growth phase is 2.5, very similar to that used by Ferguson in the IC paper, which would help account for the accuracy of the predictions made in the IC paper for ICU cases.
3: The first 7 countries have estimated R0s during the decline phase of between 0.6 and 0.8. All very similar, reflecting the similarity of lockdowns.
4. Sweden is an obvious outlier. Its estimated R0 during the growth phase is similar to the value (1.4) which I calculated in an earlier post on this site using case data. It confirms that the main reason for the poor predictions of the paper which modelled the epidemic in Sweden was its high value for R0.
5. Sweden seems to have had a significantly lower R0 before its ‘soft lockdown’ than any other European countries. However, with R0 at 1.6 Covid-19 would still have infected more than 60% of the population so some form of lockdown was necessary.
6. Sweden’s lockdown didn’t need to be as strict as other countries because its R0 in the growth phase was lower. It had only to halve the rate while other countries had to reduce theirs by up to two thirds.
7. As Sweden’s lockdown has been softer than in other countries, the effect has been less so it now has the highest value of R0, still fairly close to 1.
8. Although Sweden’s R0 is currently the highest, it’s still been reduced to a similar level to other countries, indicating that social and economic activity within Sweden is at a similar level to that of other countries.
The aim of lockdown was to reduce the deaths from Covid-19 once the disease had begun to spread widely. It’s likely that, should there be an enquiry into officialdom’s decisions on Covid-19, charts and tables like those above will be used to argue that it’s succeeded just as planned everywhere it was tried.
All lockdown countries had R0s which would have led to ~80% of the population being infected. ‘Lockdown theory’ predicts that there would be a peak in the death rate two to three weeks after lockdown (imprecision due to earlier measures and voluntary social distancing) followed by a gradual fall as R0 fell below 1. This happened in a very similar way in all countries in which lockdowns were imposed.
The apparent exception – Sweden – had a lower R0 before its ‘voluntary lockdown’. Nonetheless, had no changes been made it would have suffered a widespread epidemic. The limited lockdown which was imposed, combined with apparently voluntary social distancing reduced R0 below 1 to a slightly higher but similar level the other countries.
The evidence is that lockdowns were both necessary to stop the spread of Covid-19 before health systems were overwhelmed and worked just as those who proposed them – principally Ferguson et al – had predicted. There’s still a question mark over the IFR of Covid-19. Deaths to date (see ‘The predicted effects of lockdown”’ above) have been similar to those predicted by Ferguson et al using an IFR of 0.9% under current lockdown conditions. As this prediction relied on his model correctly estimating the number of infections and hence ICU admissions, and since this prediction has turned out to be approximately right, it’s likely that the assumed IFR isn’t far wrong.
But did it have to be this way?
Perhaps there are other explanations for the apparent success of lockdown which might become clear when the spread of Covid-19 is better analysed. A key assumption made by Ferguson in his longer term strategy is that Covid-19 maintains its pre-lockdown values of R0 and IFR so repeated lockdowns will be needed to suppress recurrences. If R0 naturally falls, as the virus evolves, or the virus has already differentially infected the population so the most infectious people have already been immunized, leaving effective her immunity, it may not be necessary to repeat the lockdown.
But if those happy circumstances don’t apply, is there an alternative strategy to lockdown which might succeed without such extensive ‘collateral damage’? To continue the euphemistic language: has any nation ever used ‘friendly fire’ to such devastating effect?
Had all countries taken Taiwan’s approach at the outset, perhaps the world would have spared itself the social and economic wretchedness which seems likely to blight the future. How did Taiwan avoid self-harm? A contributory factor is almost certainly constant alertness towards China. When your nearest neighbour has sixty times your population and has made no secret of its aim of ‘reunifying’ your country with theirs, by whatever means necessary, including force, while in the meantime trying to undermine your economic and political position and international standing, you tend to watch them closely (I was going to say ‘like a hawk’, but perhaps more like a rabbit watching a hawk).
After SARS originated in China in 2004, Taiwan made preparations and established systems to deal with future epidemics. As soon as the first cases of Covid-19 were reported in China, Taiwan didn’t prevent travel – the links of its people with China are too close to do that except as a last resort. Instead, incoming flights from China to Taiwan were screened and infected passengers quarantined. When the disease nonetheless spread beyond arriving passengers (who presumably arrived during Covid’s asymptomatic period), all such secondary cases were tracked, traced and isolated, likewise their contacts. Covid-19 never got a chance to get established. So far Taiwan has had seven Covid-19 related deaths, in a population of 23 million.
Taiwan’s almost unique situation with respect to China underlies the preparation, vigilance and prompt decisive action which it took. Likewise Hong Kong, in some ways facing even more pressing threats than Taiwan from China, took similar measures and with similar results: four deaths in a population of 7.5 million.
It seems the best way to prepare for Covid-19 was to feel threatened by China. Covid-19 was similar enough to SARS that the systems Taiwan developed as a result of fighting the last war against a viral invasion from China worked just as planned against the newcomer. The rest of the world, lacking the paranoia which comes with being constantly menaced, was comparatively relaxed about the virus until it was so widespread that preventing its entry to a country would no longer stop it spreading further within the country. And so lockdowns followed because there seemed no alternative way to stop the virus.
Perhaps future pandemics can be prevented by all countries adopting Taiwan’s methods. Just as an individual’s immune system works best when it immediately identifies an incoming virus and eliminates it before it spreads within the organism, so nations must be constantly vigilant against inward bound viral threats and ready to stop them the instant they arrive. Otherwise, if they have the virulence of Covid-19, the only recourse may be another lockdown.
In the meantime, with Covid-19 established within countries, will ‘track-and-trace’, combined with rigorous quarantine and border controls like Taiwan’s suffice to control its spread without lockdowns? I don’t know, and nor perhaps does anyone else. We need another model…