Comment: The grave matter of Covid-19 death statistics

Published in: Longevity - mortality, UK, Covid-19

Companies: Office for National Statistics, ONS, Public Health England

One of the many bewildering aspects of the Covid-19 pandemic has been the struggle to put a number on how many people actually died from the disease.

In the UK, the government initially touted a daily toll sourced from Public Health England. But it emerged that only included people dying in hospitals, and ignoring the terrible losses happening in care homes.

The Office for National Statistics (ONS) reports weekly mortality figures, that actuaries quickly analysed to produce an estimate for “excess deaths” i.e. how many more people were dying compared with the expected average.

The actuaries said this was more reliable way of measuring the impact of Covid-19 – which it was, until a few weeks ago, when the “excess deaths” figure turned negative. (Happily, mortality rates in the UK are now lower than expected, perhaps a side effect of the lockdown, or just a resumption of the low-mortality trend observed in January and February.)

But people are still dying of Covid-19, so perhaps we need to consider the government’s figures again?

Well, not so fast. It turns out these might have been an over-exaggeration because people who had been tested positive, but had since recovered, were still being counted as Covid-19 deaths.

Edward Morgan, senior research officer at the Centre for Ageing and Demography at the ONS, says analysis of all-cause mortality allows the organisation to examine the impact of the pandemic “not only from deaths due to Covid-19 but also excess deaths that have occurred as a result of the wider impacts of the virus on healthcare systems and society”.

That will certainly be true: future estimates of the impact of pandemics must take into account the possible responses from governments that could distort mortality trends – whether positive or negative.

But anyone trying to actually use a figure for Covid-19 deaths will just have to include enormous error margins, for now.

Cintia Cheong