One year on

It is almost exactly a year since my first COVID-19 blog post was published here on 13th March. So here some thoughts on the last pandemic year.

It has been a very mixed year: traumatic – watching case and death numbers climb, experiencing illness and deaths of friends; challenging – lockdowns, seeing their impact on friends and colleagues; fascinating – experiencing the outbreak as a modeller and statistician.

Having spent the last 12 months working almost exclusively on the epidemic and having written eleven The Conversation pieces and many blog posts, I have been thinking about the wrong and right things I said and wrote.

I got spectacularly wrong the speed at which the UK – and other countries – would proceed with vaccination. I expected the roll-out to be slow, with logistics of delivering the mass vaccination while maintaining social distancing, as well as vaccine hesitancy as main problems.

Israel, UK, and other countries have shown that it is possible to vaccinate at a scale and speed I would not have thought possible. Yes, there have been hiccups and delays – I still don’t understand the low uptake in Germany – but overall this is an impressive result.

Speaking about vaccines, I actually expected more side effects, including fatalities. As it has turned out, the available vaccines are indeed remarkably safe at least in the short term.

I initially got masks wrong. I imagined that a cotton mask would not provide efficient filtration. Although the mode of spread is still debated, with a non-negligible potential of air-borne transmission, there is very good evidence that masks help to suppress spread.

I underestimated the potential of SARS-CoV-2 to mutate. Early in 2020 I followed the then prevailing opinion that coronaviruses have lower potential for producing highly infectious strains. Unfortunately, the experience of winter months of 2020/21 has shown how wrong this was.

What I got right? I would claim three interlinked things. Firstly, even from early on, I have been convinced that we are in for a long fight with the virus. Working with pests and pathogens has taught me that they are notoriously difficult to eliminate and eradicate.

I have been very impressed with how China, Australia, New Zealand, Taiwan and Vietnam have so far managed to almost eliminate the virus. Their’s has been a high cost of a strict, sometimes brutal, clamp down on social interactions, as well as the lack of international travel.

As long as the virus exists elsewhere, the #ZeroCovid situation is unstable, as the Scotland summer 2020 example clearly demonstrates. So the #ZeroCovid strategy needs to be maintained. I do not know how much will there is in societies to do so.

Secondly, that it is only by building trust and coordination across different levels of society that we can stop the epidemic. Very early on I realised that the economic and social inequalities are going to have a huge impact on the spread and persistence of the disease.

Thirdly, the year has put me in a battle with the “herd immunity”-Barrington Declaration proponents. I still think that #ZeroCovid was the best strategy for containing the disease – and possibly still is.

#ZeroCovid approach consists of a rapid, deep lockdown, long enough to reduce the numbers, followed up by an efficient test-and-trace programme backed up with incentives to quarantine, and efficient border control.

I am actually amazed at how well our 2012 paper predicted the pandemic. Our model is very simple and we only looked at individual decisions rather than governmental ones. But we predicted that the #ZeroCovid strategy is economically optimal for highly infectious diseases.

However, this strategy borders on another one of continuing lockdown-release cycles, R constantly close to 1, and a long, expensive, unsuccessful epidemic. By the way, this paper has now been cited by 134 other studies (Google Scholar), 98 citations in 2020-21.

As we approach a third (or is it a fourth?) lockdown, I am painfully aware that I should have been much more vocal about #ZeroCovid throughout 2020-21. Maybe it is not too late…

On herd immunity again

I have been having e-mail and Twitter discussions about the “herd immunity threshold” (HIT) and what it means for policy. Typically, the conversations centre around four items: (i) the number of people who have been infected with the virus and are therefore potentially immune to the disease (these two things are not the same). We seem to be arriving at significantly different numbers; it would have been an academic discussion was it not for the claim that (ii) we are nearing the HIT, (iii) the drop in new cases is due to a large proportion of immune individuals rather than lockdown, and hence (iv) we can all relax and return to “normal”. There is a variation on these points, suggesting that even if we are not close to HIT, we should simply let the disease run through the population while the vulnerable are “protected” by vaccination). I will try to address these four points briefly below.

(i) With the two large waves of COVID-19, we clearly have built up some “natural” immunity – by this, I do not mean cross-immunity from other coronaviruses or BCG vaccines – but immunity from having COVID-19. The question is how much. A number of studies point to the levels being not higher than 20-30% in the general population; the evidence comes from antibody testing as well as model outcomes. These numbers can be higher locally, for example, some evidence points to 50% in Delhi and maybe as much as 70% in Manaus (although this last claim has proven controversial).

The levels of antibodies in the UK are between 10-20% and the models which “reconstruct” the prevalence predict about 15-30%.

(ii) There is quite a lot of renewed interest in the estimation of the “herd immunity threshold” (HIT). Simple models predict 60-70% based on R=3 as observed in Wuhan in the early stages of the epidemic; for the new strains this might be higher, but then we never observed them without any non-pharmaceutical control measures.

We, modellers, do know this is an overestimate, as it does not include various sources of heterogeneity. The question, however, is by how much it is overestimated. Some researchers originally came up with 20%, although they have been revising the numbers (I think ca 30% is the latest estimate); there are very solid papers that point out errors in this analysis. The problem is that each model makes assumptions that often are very difficult to validate.

(iii) I found the arguments by the “low HIT” camp unconvincing and so I think we are still some way from HIT, perhaps except some locations (some places in India come to mind).

As I understand it, Manaus is a good example of the need to be cautious. The antibodies estimates in Autumn 2020 found about 70% population positive and the researchers claimed that the first wave stopped because of the HIT. Unfortunately, Manaus then experienced the “second wave” of disease, suggesting that either the antibodies levels were lower, or the protection was overcome.

One thing to realise is that the models predict that once we start approaching HIT when the infection levels are high (as now), the reduction should be relatively slow. The “threshold” is not really that dramatic.

(iv) I have the main issue with people who claim that we are already close to or at the HIT and so we can all relax, drop all precautions and just return back to “normal”. This is a recipe for disaster if – as I believe – we are not past the HIT and the consequences are terrifying. Why? It is because I do not believe the “third” lockdown will work – people will stop complying and we will have a wave of the infection going through the population again, with huge loss of life and long-COVID.

Polsat TV – cd

Brief English summary: Four widely-accepted models are used to estimate the “true” number of infections from COVID-19. The cumulative number for Poland in February 2021 is 26% (ICL) and 14% (IHME); for comparison, the UK has 37% and 16%.

Po napisaniu poprzedniego blogu, poszukałem wyników modeli dla Polski, które umożliwiają ocenę “przwdziwej” ilości zachorowań. Szczegóły są na stronie internetowej:

ICL to Imperial College London (UK)
IHME to The Institute for Health Metrics and Evaluation, University of Washington (USA)
YYG to Youyang Gu, matematyk i informatyk z USA
LSHTM to The London School of Hygiene & Tropical Medicine (UK)

Używając wyników z tych powszechnie zaakceptowanych modeli, można wyliczyć, że do tej pory 26% (ICL) lub 14% (IHME) populacji przeszło przez chorobę w okresie do lutego 2021. Wyniki podane przez YYG i LSHTM nie sięgają do 2021, ale w sierpniu (0.7%) i pazdzierniku (3.8%) były zbliżone do ICL i IHME.

Nawet przyjmując pesymistyczne wyniki z modelu Imperial College London, mamy w Polsce 26% osób potencjalnie odpornych.