Population density and COVID case rates

Public Health England is releasing daily data on confirmed cases of COVID-19 in England, broken down by region. The data gives a simple running total of cases by upper-tier local authority (another word for certain types of local government). Can plotting this data give us any additional insights?

The map above takes the COVID-19 cases data, combines it with some statistics on population size for each area, and generates a continuous scale for cases per 100,000 people. Doing this helps to control for the fact that each of the areas plotted above varies wildly in the number of occupants. It isn’t fair to compare case numbers in a London Borough with a rural county, so transforming COVID-19 cases from a raw number into a rate makes this map far more useful. It’s called a univariate map, because it plots just one variable.

There’s no mistaking that London is the epicenter of the pandemic in England. All of its boroughs are among the top areas for COVID-19 cases per 100,000 inhabitants.

Hampshire in the south and Cumbria in the north west are two large rural counties that have reported a high case rate. Birmingham and surrounding areas in the midlands have high case rates too as does Sheffield.

Of course, this can reflect testing practices in different areas, particularly the extent to which testing for COVID-19 is actually taking place. Areas with more testing will report more cases, something that must be kept in mind with all these figures.

This gives an idea of distribution, but what about the spread of the virus? Densely populated areas offer more vectors of spread for coronavirus, whereas in sparsely populated areas we might expect to see a lower case rate.

Unsurprisingly, central London is the most densely populated area in England, with other major cities visible as darker areas.

What does the data look like when we combine information about population density with COVID-19 case rate in the same map?

This is a bivariate map, because it brings together two variables into the same map.

If case rate of COVID-19 is related to population density, we should expect to see lots of dark violet areas, indicating high case rate and high population density. This appears to be true in London and some other major cities, but for other areas the picture is more complicated.

Outside of London, most of the smaller areas on the map represent metropolitan districts (cities or large towns to you and me). Many of these cities fall into the medium blue category, for densely populated areas with low case rate. Its possible that the virus is not spreading in all major cities in the same way as it has in London, although the extent of testing in different areas is unknown.

Kent, Essex, and most counties in the midlands are sparsely populated, with a medium case rate. Some of these counties are commuter areas for London or Birmingham, where case rate is much higher. Hampshire and Cumbria remain prominent on this map, being sparsely populated but with high case rates. Hampshire is a particularly strange case. Its two major cities, Southampton and Portsmouth, are reported separately, and both are more blue than red, indicating lower case rates despite a high population density.

The east, including large rural counties like Cambridgeshire, Norfolk, and Lincolnshire, have seen relatively low case rates. So too has most of the West Country.

The figure below is another way of looking at the same data outside of a map. The colours correspond in the same way- dark violet for densely populated areas with high case rates.

It continues to be obvious how much of an outlier Cumbria is, but another outlier has appeared. Kingston-upon-Hull is a city of 250,000 people, but it has the lowest case rate of any region in England. Again, this could reflect a lack of testing.

At this time, larger case rates appear to be focused on London, Birmingham, the corridor between them and their commuter belts, but how has this pattern developed? In the next article I’ll look at how this picture has changed since the first confirmed cases.

All analysis and plotting using R.