Last night the UK Prime Minister said that those who could not work from home, like those in construction and manufacturing should go to work today, maintaining social distancing, and avoiding public transport if possible. Primary schools may begin reopening in June, as may some shops, and some of the hospitality industry may reopen in July. The advice was conditional: if the reproductive rate goes above 1 (it is estimated to be between 0.5 and 0.9) then these relaxations may have to be reversed.
Bizarrely, in my view, he added: “it will soon be the time… to impose quarantine on people coming into this country by air”. If that had been implemented in January the UK would not have had an epidemic. For some reason the argument he gave was that, because good progress had been made in reducing R by shutting down society, it was now appropriate to deal with the supply of the virus and in due course, without any particular urgency, turn off the tap that set the whole country afire.
So, it is not an end to lockdown, nor the beginning of the end, but possibly the end of the beginning.
Lots of ideas are rattling around the country, and it is salutary to note some of them, since we are getting so much history now. Decades when nothing happens, and then days when decades happen.
There is considerable scepticism about the Imperial College model, which is seen as poorly documented, poorly configured in computing terms (bad architecture). There is plain scepticism about the accuracy of any of Prof Ferguson’s predictions, because he has a history of wildly exaggerated death rates. Or that is the claim against him. More likely, each of his simulations involved several predictions: best case, most likely case, worst case and both he and his critics can pick those which best suit their story. What is Prof Ferguson’s Brier score (mean squared difference between predictions and actual outcomes)? How many predictions does he make per simulation? Would he get to super-forecaster levels? At the moment it does not seem likely. Interesting, however, that his predictions are mutative: even if wrong, they have had a big influence on policy, and so alter the outcomes which serve as the test of predictive accuracy.
He did not do well with the foot and mouth outbreak. Long before I took up blogging I wrote to the relevant Ministry to ask for their justification for their slaughter of cow herds, and was fobbed off for months until on the third attempt they sent me what they claimed was their justification: not the work of Ferguson but a calculation by another professor. I contacted him and he replied that his note to the ministry was a back of an envelope calculation, and no basis for any policy at all. Foot and mouth disease does not lead to appreciable drops in animal fitness, and the meat is still fit to eat, so the whole episode remains something of a mystery. I digress.
What should one require of all model simulations on which public policy may be based? At a minimum, it seems essential to have an introductory page giving all the assumptions in plain language, and in a standard agreed format. Then, a second page explaining the basic structure of the computer program. Then, publication of the actual code for inspection and expert testing.
It is extraordinary that one of the most important events in current economic and social history was based on the unexamined workings of a computer model. It appears that politicians believed the numbers because they were printed out by a computer. An accountant friend was once asked to audit a very smart perfume shop in Mayfair, London. It was a palace of marble and chrome, every saleswoman a beauty, and the accounts beautifully printed out by computer, which was novel at that time. He asked himself: If I owned this shop and was trying to steal money, how would I do it? Suspicious, but at a loss where to begin, he spent some days adding up all the entries with his own calculator. The computer totals had been falsified, but very neatly, and they were skimming money out of the business.
Was lockdown a waste of time, because even if it prevented prompt deaths, herd immunity is lowered, leaving us open to a much more dangerous second wave this winter? This is an unpleasant Is it 66% or a far lower figure? If most super-spreaders are already immune, very likely because they were most likely to catch the virus in the first place, then the herd could be safe at a lower threshold. In the UK at least, the picture is still unclear.
On a far brighter note, there is now more clarity about how the infection spreads. As a result of a series of tweets (a chorus of tweets?) Dr Muge Cevik of St Andrews University deserves two medals: one for not having a computer model; the other for looking in detail at 14 studies of close contacts of Covid cases, showing how many infected people go on to infect how many others, and how those rates differ between sustained indoor settings and more casual outdoor ones. Apart from confirming the age gradient in vulnerability, she shows very clearly that sustained exposure in an enclosed space is the greatest vector of infection (houses, offices, public transport). Casual interactions outdoors are far less risky. Looks like droplets, not aerosols, are the main vector. The advice would be: stand apart, wear masks, wash hands, and reconfigure public indoor settings to reduce all respiratory disorders in future.
While the infectious inoculum required for infection is unknown, these studies indicate that close & prolonged contact is required for #COVID19 transmission. The risk is highest in enclosed environments; household, long-term care facilities and public transport.
High infection rates seen in household, friend & family gatherings, transport suggest that closed contacts in congregation is likely the key driver of productive transmission. Casual, short interactions are not the main driver of the epidemic though keep social distancing!
Increased rates of infection seen in enclosed & connected environments is in keeping with high infection rates seen in megacities, deprived areas, shelters. A recent preprint demonstrates that #COVID19 epidemic intensity is strongly shaped by crowding
Although limited, these studies so far indicate that susceptibility to infection increases with age (highest >60y) and growing evidence suggests children are less susceptible, are infrequently responsible for household transmission, are not the main drivers of this epidemic.
Finally, these studies indicate that most transmission is caused by close contact with a symptomatic case, highest risk within first 5d of symptoms. To note: this preprint suggests that most infections are not asymptomatic during infection
In conclusion, contact tracing data is crucial to understand real transmission dynamics. Cautionary note: This data & interpretation is based on the available evidence as of May 4th. Our understanding might change based on community testing/lifting lockdown measures.
Addendum: While we have limited data, similar high risk transmission pattern could be seen in other crowded & connected indoor environments such as crowded office spaces, other workplace environment, packed restaurants/cafes, cramped apartment buildings etc.
Conclusion 2: (a) we need to redesign our living/working spaces & rethink how to provide better, ventilated living/working environment for those who live in deprived & cramped areas; (b) avoid close, sustained contact indoors & in public transport, & maintain personal hygiene.
The probability of getting infected when you are in close contact with an infected person tells us a lot, and puts things into perspective. in household, this risk is about 15-20% (so 1 in 5 chance), but in crowded closed places this can go up to 40% (super spreading events) but casual short interactions are far less risky.
There is further opinion on infectivity here: https://www.erinbromage.com/post/the-risks-know-them-avoid-them
From Brazil comes a delightful piece of work.
The authors note that, given the fact the virus originated in Wuhan but is now found world-wide, it must have got there by some means. They correlate country levels of infection with the number of airplane flights each country receives, and find that the main explanation is air traffic, which accounts for 45% of the variance. Socio-economic and climatic factors are less powerful. Who would have thought it?
Moral: when an epidemic is in the offing, close down airports, first banning flights from those countries or parts of the country where the infection originated, then to those countries which have air links to the originating country, but best of all, just close airports. Then, once you have all the facilities in place, accept a restricted number of flights for those people you may deem necessary to let in, but test and quarantine (in central facilities) as you see fit.
Ban planes, muffle sneezes, clean up droplets.
There is a simple psychological point about epidemics: it is best to take draconian actions before they are necessary. Better to lose the airline industry than the whole of industry.