Almost 10,000 excess deaths, not recorded as COVID-related
Early results from a study led by University of Manchester data scientists has shown that of the 47,243 excess deaths in England and Wales in the first nine weeks of the COVID-19 pandemic, 9,948 – over a fifth – were not recorded as COVID-19 related.
The excess deaths were mainly in older people and in many cases likely to be directly linked to COVID-19 despite not being diagnosed as such. But, say the team, the excess deaths also include deaths driven by the inability or reluctance of people to access health services for other health needs.
The team, using Office for National Statistics (ONS) weekly mortality data, also showed the highest rates of excess deaths, linked to COVID-19 directly or indirectly, were in the West Midlands: 1,531 deaths or 26 per 100,000 people.
Although many more male deaths have been linked to COVID-19 across all age groups, female non-COVID-19 excess deaths showed similar patterns to male deaths, and female deaths outnumbered male deaths in the most vulnerable 85+ age group. COVID-19 deaths in women may, therefore, have been underreported argue the team
The researchers, using they argue more robust approaches than recent ONS analyses, analysed a time series model of weekly data from 1 Jan 2010, accounting for seasonality and any potential changes in the overall mortality trend.
“We were able to quantify excess deaths which were not recorded as being COVID-19 related, when perhaps they should have been, by analysing weekly all-cause mortality data from 1 Jan 2010 and including COVID-19 linked deaths from 7 March 2020,” said Professor Evan Kontopantelis from The University of Manchester.
“There is great seasonality in mortality data, with more deaths occurring during the winter months. Using this long time series we estimated how many deaths were expected from 7 March 2020, for each population group and region, and compared this against what was actually observed.”
The regions least affected were Wales with 1,757 excess deaths (including COVID-19 deaths) and the North East with 2,239 deaths. These regions were also the least affected when the population size was taken into account.
We identified a large number of excess deaths which have not been linked to COVID-19. Some of these probably should have been linked, while the rest are caused indirectly by the response to the virus. ONS reports of COVID-19 related deaths are based on clinical judgments, which in some cases will have been made without positive testing, which might explain why so many COVID-19 deaths went unrecorded.
Excess death rates overall started to increase after 20 March and peaked between 17 and 24 April, with the highest weekly peaks in London (up to 26 per 100,000 people), the West Midlands (up to 24) and the North West (up to 24).
The models developed by the team, based on historical mortality trends, illustrate the scale of the COVID-19 pandemic and its direct and indirect impacts on population health.
Professor Kontopantelis said: “We identified a large number of excess deaths which have not been linked to COVID-19. Some of these probably should have been linked, while the rest are caused indirectly by the response to the virus.
“ONS reports of COVID-19 related deaths are based on clinical judgments, which in some cases will have been made without positive testing, which might explain why so many COVID-19 deaths went unrecorded.
“Deaths during the COVID-19 pandemic have resulted directly from infection and exacerbation of other diseases. But there has also been an indirect effect from deferment of care for other conditions.
“Our figures are likely to underestimate the impact of the virus, as prior to the pandemic mortality rates for 2020 were relatively low, particularly compared with years with high influenza activity.”
He added: “We need to be cautious if we are to predict longer-term impacts, but given the sudden re-prioritisation of services, it is likely a second wave of non-COVID-19 related need is currently building in the community.
“And that could overwhelm parts of the NHS and communities with high baseline levels of morbidity. This should be recognised in planning post-pandemic health and social care services.”
A pre-printed version of the paper is available here. It is not yet peer-reviewed, but the authors felt it was in the public interest to release the study in its early form