25 May 2020 ; tagged in: data

Analysing the Covid-19 data

A summary and extrapolations of the UK's Covid-19 data

Analysing the Covid-19 data

Most recent update:

As with many people, I've been looking at the Covid-19 data. I've analysed the number of deaths, as the number of reported cases is strongly influenced by the number of tests performed. These figures will be updated daily, as more data comes in.


There have been deaths reported in the UK, up to .

Based on ONS data for excess deaths, I estimate the true figure of Covid-19 deaths to be .

Reported deaths, by country

This graph shows just the number of deaths reported in each country. It will be an under-estimate of the true impact of Covid-19, as many people who die from the disease will do so without a formal diagnosis.

As the epidemic accelerated in different countries at different times, the horizontal scale in this graph is number days since ten Covid-19 deaths were reported in each country.

Total deaths in selected European countries

Excess deaths in the UK

This data comes from the ONS, NISRA, and NRS, so covers the whole of the UK (see data sources below).

Excess deaths in the UK

From week 12 (week ending 20 March 2020) until week ending , there were approximately excess deaths, based on the average for the previous five years. In that period, the UK reported Covid-19 deaths. Since then, the number of deaths has been in line with the expected numbers based on previous years.

The UK has reported deaths up to . Using the excess deaths up to and reported Covid deaths since, I infer that there have been total deaths so far.

Deaths per day

These figures are based on a seven-day moving average. (Note that there have been several large corrections to death numbers, particularly for Spain and the UK.)

Here are the deaths across Europe in the last 30 days (again, 7 day moving averages).

Prediction of future deaths

Previously, I was using Italy's deaths data to predict the UK's deaths data. This worked when both countries' trends of deaths were falling or constant, as they were until September.

As of mid-September, with cases rising in both countries at around the same time, I can't use Italian data to predict the UK's future deaths.

Hospital admissions

The number of deaths, new cases, and hospital admissions provide different views into the prevalence of Covid in the population. It seems that people are admitted to hospital about ten days after the onset of symptoms, and die fifteen days after admission (typical times for severe cases). I can plot the numbers of these three metrics, shifted by the appropriate delays, allowing comparison of a death today with an admission fifteen days ago and a new case twenty-five days ago.

These are all given in arbitrary units, normalised to 1 on 15 May. This is because the numbers of cases, admissions, and deaths are all very different.

This graph shows that testing started becoming effective around the beginning of July (twenty-five days after the date shown) and that a decreasing fraction of hospital admissions lead to death. This is likely  a combination of better understanding of how to treat Covid, and fewer old people having the disease.

Cases and lockdown

With a second wave of Covid possibly starting in several European countries, it's useful to compare the number of cases across countries. Note that the testing regimes and capacities are much larger in the autumn than they were in the spring and summer. Note that this graph using the number of new cases against dates, not against days since ten deaths.

Testing effectiveness

Rising numbers of Covid cases could be because of increased prevalence of the disease, or just an increase in the number of tests being performed.

One way to tell the difference is to look at the fraction of tests that give positive results. If the fraction of positive tests increases, it indicates that the true number of cases is also increasing; if the fraction of positive tests decreases, it indicates that the testing is being applied more widely to a population with a non-increasing case load.


European data on deaths comes from the European Centre for Disease Prevention and Control.

Population data on excess deaths comes from:

Hospital and testing data from the Gov.UK Coronavirus data portal.

You can download the source code that generated these figures.