Statistics explained: how to make sense of ‘excess’ deaths
How many people have died because of the pandemic? How does this vary across countries? These are two of the most common questions I get asked and yet they are remarkably difficult to answer.
We could start by looking at the number of Covid deaths listed on a website and compare countries by Covid deaths per million population. But this assumes the way countries record a death as “Covid” is consistent and ignores any deaths caused by lockdown measures and disruption to health services. It’s fairer to look at what has happened to the total number of deaths.
Some people would have died even without the pandemic; we are interested in the excess compared with that number. The problem then becomes estimating the deaths we would have expected anyway. Many methods are used: the Office for National Statistics (ONS) makes a comparison with the average over the last five years, while Public Health England uses a more complex adjustment for changes in population and trends in mortality. They arrive at similar conclusions.
From ONS data, we can calculate that between September and the end of 2020 there were around 21,000 excess deaths in England and Wales. Yet there were 29,000 deaths with Covid on the death certificate, so there has been a substantial deficit of around 8,000 nonCovid deaths. We are, then, faced with a statistical challenge – how could this happen?
First, around half of this 8,000 will be people whose primary cause of death was something else: they died “with”, rather than”‘from”, Covid. Second, there has been almost no flu. But sadly, we are seeing the absence of the many elderly people who were among the 59,000 excess deaths between March and June, but would otherwise have survived until later in the year.
One consequence is that, once the vaccines start to work, we may even find the excess in deaths becomes a deficit – a huge relief, but this could make it difficult to argue for continued restrictions.
David Spiegelhalter is chair of the Winton Centre for Risk and Evidence Communication at Cambridge