web analytics

How important is COVID-19 epidemic?

I have recently been seeing a number of posts on Facebook still questioning the size and importance of the COVID-19 outbreak. Although it has now been going on for over 4 months, there are still people who make some of the following statements:

  1. There is nothing really happening as it is not worse than usual with seasonal flu;
  2. There is an epidemic but the level of deaths is nothing compared to other diseases;
  3. There is an epidemic and there is a lot of excess deaths, but the strategy chosen to stop the outbreaks causes more deaths.

As this blog is about statistics, I will use the publicly available data for England and Wales and for Scotland to address these points. The numbers are official records of deaths as collected by registrars and so are not covering the most recent weeks. On the other hand, they are very reliable, as every death in the UK needs to be reported. I will be more cautious about the associated cause of death, as this potentially involves a medical interpretation or tests and is, therefore, more open to criticism.

By using centrally collected data I will avoid a problem of self-selection which makes many Facebook posts so unreliable. Typically, in these posts, we are presented with an opinion of or an interview with a medical doctor or a nurse who talks about their own experience in a particular location. We need to understand that the outbreak levels in large countries like China or the US can be extremely uneven so that we can get a very big problem in New York City and practically none in the rural location far from a big city. Similarly, an experience of somebody in London might be very different from the one from the Scottish Highlands or the Outer Hebrides.

Selecting either one or another and saying this applies to the whole country is disingenuous at best and a dangerous lie at worst. This is why I just want to stick to overall official data. But if you think that the Registrar data are biased and incorrect, you can stop reading now and forget about my arguments.

Let’s look at the first claim. To address it, I will just use the number of deaths recorded per week in the UK. The data are publicly available and there is a nice visualisation of them available here. The graph below shows the number of deaths in each week of 2020 as officially reported, compared with the number of deaths in that week over the last 5 years.

The deaths in previous years follow the ‘usual’ pattern, higher over wintry months, going down as temperatures go up. The numbers are between 10 and 15 thousand. The population of England is about 56 million and of Wales about 3 million, together 60 million (Scotland adds another 6 million). The resulting death rate is 17-25 per 100,000 people, about right for a country like the UK. A lot of winter deaths are associated with seasonal flu, but this goes away in spring which is why it drops down in March.

Up until 17-24th March the deaths were actually lower than the average over the last 5 years, but within a usual variation. However, there is clearly something unusual about the deaths in the weeks starting from the week of 27th March. The epidemic in the UK started in early March, but it takes several weeks for the first critical cases to succumb to the disease. What we see in the data is the massive increase in late March and early April is the result of COVID-19 epidemic. As of 17th April, the last week for which we have data (it takes about 4 days for the death to be registered), the number of excess deaths is more than double the ‘normal’ number and getting worse.

Note that this includes all deaths, whether COVID-19 related or not (points 2 and 3 above); I will address this in a separate post.

While we are looking at the data from England and Wales, let’s see what the regional differences are; shown here is the number of weekly deaths per 100,000 inhabitants.

Broadly speaking, London is hit worse. South East which is quite affluent, had the lowest death rate before the COVID-19 outside of London, but was then hit quite badly. Wales and North East, which are a mixture of rural communities and relatively deprived post-industrial towns were hit relatively less.

Wales is particularly interesting, as the level of deaths is here similar to the one seen in January, possibly associated with seasonal flu. But March is not a season of flu and so the excess of 7 deaths per 100,000 inhabitants is very unusual for the time of the year. I will show Scottish data in a later post, as I am waiting for the April data to be released.

To summarise in response to point 1:

  1. There is nothing really happening as it is not worse than usual with seasonal flu;

the answer is that indeed there is actually something happening and is not usual for this time of the year. While for rural communities we could possibly currently claim similar levels of death as at the seasonal influenza peak, firstly, this is not seasonal influenza season, and secondly, urban and affluent places like London or South East – or New York – are hit even more.

In the subsequent posts I will look at other points.

Levels of immunity

One of the key questions underlying the future course of the coronavirus epidemic is the level of immunity in the population. In other words, it is the proportion of people who have been through the disease, either with or without symptoms, who are now immune to it, at least for some time. This proportion is very important, as it determines the famous (or infamous) “herd immunity”.

There is a huge effort now on trying to establish this level. It is not an easy task, as there is no simple test that can tell us that we had the disease and that we are immune. There are also logistic difficulties in how we can measure this level across the whole population. There are different methods and the most popular involves taking blood samples and testing for the presence of antibodies. These are chemicals that the body uses to fight off the disease, and if you have had COVID-19, you would have the antibodies in your blood (this is all very simplified here).

There is now an increasing number of such studies available and a couple landed in my inbox this morning. There is a study from a single hospital at in Wuhan; more studies are summarised by an article on Bloomberg web site (which otherwise concentrates on the fatality rate, about which I will write a different post).

The study in Wuhan, which was the epicentre of the pandemic, suggests about 10%. The city of New York shows 21.2% while the State of New York is about 13.9%. In a study from Germany, we are seeing the levels of 14% and from Switzerland, 5.5%. The important thing about these studies is that they attempt to test more or less a representative group of the population so that we can get an idea of what actually is happening outside the hospitals and care homes.

There are two lessons from these very early studies. Firstly, the numbers point to a substantial epidemic. About 20 million people live in the City of New York; assuming a conservative proportion of 13% holds for the whole population (i.e. that the study was representative), we get 2.6 million people who were so far infected. The whole New York State currently has 300,000 reported cases, so the under-reporting is about 8:1. In other words, for every reported case, we see 7 unreported ones. This is actually not too dissimilar to such diseases as flu (both pandemic in 2009-10 and seasonal) and measles before the vaccination.

Secondly, these numbers are very low for “herd immunity”. Given what we know about the rate of spread of the SARS-CoV-2 virus without the social distancing measures, we need about 50-70% of the population to be immune to stop the spread and re-emergence of the disease once the lockdown is removed.

Why is it so important? The epidemiological theory can be used to explain the relationship between social distancing, immunity levels, and the reproduction rate which measures how fast the virus spreads in the current situation. These three factors are closely linked. The more social distancing measures we apply, the lower the rate is and the lower the “herd immunity” is required to stop the disease from spreading. However, if we relax the distancing, the rate shoots up and so does the “herd immunity” level.

At the moment, most countries implement some forms of social distancing, reducing the basic reproduction rate to close to or below 1. If the rate is below 1, the disease will slowly die out regardless of the immunity levels, but only a small proportion of the population will become immune. But even if the rate is only slightly above 1, once we reach the appropriate immunity level, the disease will start to disappear. New Zealand is now close to local eradication of the virus by having reduced the rate very early and keeping it below 1. South Korea uses a different method but the underlying mechanism is similar, i.e. the reduction of the rate. In contrast, Sweden appears to have reduced the rate much less but is hoping to reach the “herd immunity” level corresponding to its current social distancing measure.

However, all these countries at some point would need to relax the lock-down measures. This will result in an increase in the effective reproduction ratio. Countries like New Zealand hope that by having reduced the local number of cases to zero or almost zero and by having implemented a stringent border control, they can cope with the increased rate. Such a strategy is risky, as shown by Singapore (and mainland China) which after a successful eradication campaign are now seeing an increase in imported cases. Another strategy is to rely on “herd immunity” to reduce the potential of the disease to spread, and Sweden appears to be following this course. This is risky as well, as it potentially leads to a lot of otherwise avoidable deaths. The UK and the US are still far from any of these considerations.

The overall message is that unfortunately, the social distancing measures will be with us for quite a while, but given the underreporting, perhaps not as long as we might have feared.

New article

I have just published a new article on The Conversation, titled, Coronavirus: how to use a vaccine when it becomes available. In it, I say:

As we are waiting for the coronavirus vaccine, it is important that we consider now how it can be used in the most cost-effective and publicly acceptable way. We need to know how to balance the demand for herd immunity with the protection of individual rights.

To me, this is the key message from this article. Nothing will work if people are not on board.