Excess Deaths Changes to Methodology
Date published:
Today ONS published a new UK-wide methodology for estimating expected deaths, used in the calculation of excess deaths.
Excess deaths
‘Excess deaths’ is the term used to describe how the number of deaths observed (registered) during a certain period differs from the number of deaths that might be expected (also known as the baseline) during that period, in ‘usual’ circumstances. Although named ‘excess’, the difference can be in either direction, that is, there could be more of fewer deaths than would be expected in a defined period.
Excess deaths = Observed deaths - Expected deaths
Change in Methodology
Previously the UK statistical agencies (NISRA, NRA and ONS) used a relatively simple method to compare the number of deaths registered in the current reference period with an expected number based on a five-year average of the count of deaths registered during the same period in previous years.
The increased interest in deaths data, and excess deaths specifically, during the pandemic, showed that organisations and government departments across the UK were using a number of different methods to calculate an estimate of expected deaths. It highlighted the need for a single, harmonised approach to facilitate comparisons and the production of a single UK estimate. A cross-UK working group, led by ONS, was set up to take this work forward over the last year.
This morning, ONS published a methodology paper outlining the new, agreed model for estimating excess deaths along with data tables giving a back series of excess deaths data based on the new approach. ONS have also published a blog detailing the background to this work.
Northern Ireland Excess deaths
To supplement the ONS documents, NISRA has produced a summary information paper on the change in methodology and the impact to NI excess deaths specifically. In addition, for user convenience, the NI specific data published by ONS has been extracted into separate tables and are available alongside the information paper.