REVIEW - unconvincing papers on effects of lockdown
FINALLY - and it has taken four months - some researchers have attempted to evaluate the changes from adopting lockdown measures. It is absolutely astonishing the world's governments instituted such drastic measures so rapidly and costing trillions of dollars in total without any kind of evidence or indication it would work or was worth the economic destruction - but they did, almost without exception and generally without argument until the public tired of such extreme changes to their way of life.
By comparison, for months the same countries were all extremely reluctant to put even simple low-cost highly effective measures like closing borders into action - and they ended up having to do it anyway. It does go to show that when governments whine they have no money to undertake important economic or social initiatives - they are full of it. It's all a case of political will.
At any rate, we now have two articles in Nature from 8 June. As a wit once remarked - just because it's in Nature doesn't mean it's wrong. These papers always have about 15 authors and it is hard to believe most of them contributed in any significant way - but it gives the whole lab a high-profile publication.
The first 'accelerated article' is Flaxman et al from the infamous Imperial College at London, whose models were so influential on world governments - but lacked scientific basis or even programming skill. As usual, the infection fatality rate is taken as a random guess. From the guess however they calculate that across 11 European countries, 12-15 million individuals 'must have' been infected up till 4th May, representing 3-4% of the population (the official number was 3.6 million so they are setting undetected cases at about 3:1, on the high side and considerably above what random sample testing has shown).
The basic idea is they assume a changed R0 for different interventions. They assume nothing much for social distancing, but a massive drop from 3-4 down to 0.5 to 0.7 for complete lockdown a couple of days later - and then see how that resembles what panned out. Yes it does look rather like the final result in those countries. But sorry, not good enough, just dressing up prejudices in model form. Guys, it could have been any of the interventions, not just your favourite. But published anyway, so must be right.
According to the Washington Post, this article says that lockdowns "saved about 3.1 million lives in 11 European countries, including 500,000 in the United Kingdom, and dropped infection rates by an average of 82%", though I can't see this anywhere in the article.
The second article, Hsiang et al from UC Berkeley, was actually sent on 22 March before this blog was started - so it has taken 11 weeks to get to pre-publication (you may have been wondering why I choose to publish here). They use econometric methods rather than a model, which in theory ought to be more robust.
They correctly state the natural spread of infections exhibits almost perfect exponential growth (unlike one Nobel laureate who didn't), estimating that COVID accelerates at 38% per day without intervention (obvious in Western countries). They looked at hundreds of sub-national regions, working backward from death data to estimate infections, and calculated change following a raft of 1717 different interventions.
They "compute that the minimum susceptible fraction across administrative units in our sample is 72% of the total population (Cremona, Italy) and 87% of units would likely be in a regime of uninhibited exponential growth (> 95% susceptible) if policies were removed on the last date of our sample". In other words, they think that herd immunity might kick in at 87%.
They do make a reasonable estimate that the growth rate in China slowed from 33% to 13% in the second week and then to 5%. This delay in lockdown effect can be seen almost everywhere.
"We estimate that across these six countries, interventions prevented or delayed on the order of 62 million confirmed cases, corresponding to averting roughly 530 million total infections". Well - yes - you just take the difference between the herd immunity level and the confirmed level, model not necessary. However they are now assuming true infections are 10x the rate of confirmed infections, getting up towards the 25x Sweden fantasy.
However, while interventions might slow down the rate of infection, according to these authors the epidemic never stops until herd immunity is reached. What I have called 'case 1' where the infection is smothered and goes down as fast as it came up never happens - yet it was the first to be detected!
A severe credibility gap exists in all these models and it is not surprising that epidemiologists gave such bad advice.
SUMMARY While I am very happy to acknowledge good research, this is not it. They state the obvious, play with their models, put in a few values of parameters that confirm their prejudices and give the answers they want. They do not model any possibility for extinction of the virus - as actually happened with SARS, MERS and with COVID-19 in a number of lucky countries. Is this science?
At any rate, we now have two articles in Nature from 8 June. As a wit once remarked - just because it's in Nature doesn't mean it's wrong. These papers always have about 15 authors and it is hard to believe most of them contributed in any significant way - but it gives the whole lab a high-profile publication.
The first 'accelerated article' is Flaxman et al from the infamous Imperial College at London, whose models were so influential on world governments - but lacked scientific basis or even programming skill. As usual, the infection fatality rate is taken as a random guess. From the guess however they calculate that across 11 European countries, 12-15 million individuals 'must have' been infected up till 4th May, representing 3-4% of the population (the official number was 3.6 million so they are setting undetected cases at about 3:1, on the high side and considerably above what random sample testing has shown).
The basic idea is they assume a changed R0 for different interventions. They assume nothing much for social distancing, but a massive drop from 3-4 down to 0.5 to 0.7 for complete lockdown a couple of days later - and then see how that resembles what panned out. Yes it does look rather like the final result in those countries. But sorry, not good enough, just dressing up prejudices in model form. Guys, it could have been any of the interventions, not just your favourite. But published anyway, so must be right.
According to the Washington Post, this article says that lockdowns "saved about 3.1 million lives in 11 European countries, including 500,000 in the United Kingdom, and dropped infection rates by an average of 82%", though I can't see this anywhere in the article.
The second article, Hsiang et al from UC Berkeley, was actually sent on 22 March before this blog was started - so it has taken 11 weeks to get to pre-publication (you may have been wondering why I choose to publish here). They use econometric methods rather than a model, which in theory ought to be more robust.
They correctly state the natural spread of infections exhibits almost perfect exponential growth (unlike one Nobel laureate who didn't), estimating that COVID accelerates at 38% per day without intervention (obvious in Western countries). They looked at hundreds of sub-national regions, working backward from death data to estimate infections, and calculated change following a raft of 1717 different interventions.
They "compute that the minimum susceptible fraction across administrative units in our sample is 72% of the total population (Cremona, Italy) and 87% of units would likely be in a regime of uninhibited exponential growth (> 95% susceptible) if policies were removed on the last date of our sample". In other words, they think that herd immunity might kick in at 87%.
They do make a reasonable estimate that the growth rate in China slowed from 33% to 13% in the second week and then to 5%. This delay in lockdown effect can be seen almost everywhere.
"We estimate that across these six countries, interventions prevented or delayed on the order of 62 million confirmed cases, corresponding to averting roughly 530 million total infections". Well - yes - you just take the difference between the herd immunity level and the confirmed level, model not necessary. However they are now assuming true infections are 10x the rate of confirmed infections, getting up towards the 25x Sweden fantasy.
However, while interventions might slow down the rate of infection, according to these authors the epidemic never stops until herd immunity is reached. What I have called 'case 1' where the infection is smothered and goes down as fast as it came up never happens - yet it was the first to be detected!
A severe credibility gap exists in all these models and it is not surprising that epidemiologists gave such bad advice.
SUMMARY While I am very happy to acknowledge good research, this is not it. They state the obvious, play with their models, put in a few values of parameters that confirm their prejudices and give the answers they want. They do not model any possibility for extinction of the virus - as actually happened with SARS, MERS and with COVID-19 in a number of lucky countries. Is this science?
Comments
Post a Comment