Darryl Brundle, Head of Longevity at Legal & General and a member of the CMI Executive Committee, gives an overview of the different methodologies used to calculate excess deaths.
Since the start of the pandemic, there has been much greater interest in ‘excess deaths’.
These are typically positioned as the measure of how many more deaths have occurred than a benchmark, which is often taken as an indication of what might have been expected.
Underneath this simple concept are a range of methodology choices that have an impact on the level of excess deaths calculated, and therefore the resultant messages.
There is no single ‘correct’ method of calculating excess deaths. Different approaches can be useful in different circumstances.
This blog explains what the CMI means by excess deaths in its own publications and compares this to two other prominent approaches used elsewhere.
Within the weekly CMI mortality monitors we compare current mortality rates, calculated from death registrations, to 2019 mortality rates. When the CMI reports there have been excess deaths in a week, we mean that mortality rates for that week are higher than the equivalent week in 2019.
Within this calculation we age standardise, so that changes in the make-up of the population between 2019 and now do not influence the conclusion.
The world has changed significantly since 2019, so why do we think this is a useful measure?
The CMI methodology answers the question: ‘How far from pre-pandemic mortality are we?’. We are treating mortality rates observed in 2019 as a reasonable proxy for where mortality rates might have been in recent years, absent the pandemic. A data-driven explanation of this is given in the appendix of the week one 2023 mortality monitor.
What are the limitations of the CMI method?
The methodology is a comparison to a known past date, and we cannot know what mortality really would have been without the pandemic. Our approach does not allow for the changes in the longevity landscape we’ve witnessed over the last three years or the current pressures on life expectancy in the UK.
The CMI approach is not attempting to tell us what a ‘new normal’ level of mortality could be or whether that is different to pre-pandemic mortality rates.
Other methodologies are available
The Office for National Statistics (ONS) uses a different approach to the CMI. The ONS defines excess deaths as those above a five-year average. For 2023, the ONS will compare deaths registered in each week to the average deaths registered in the equivalent weeks of 2017, 2018, 2019, 2021, and 2022 (ie excluding 2020).
Within the ONS’s weekly provisional deaths files, they compare absolute number of deaths with no allowance for changes in population, although in other outputs, such as their monthly mortality analysis, they do calculate age-standardised mortality rates.
All else being equal, the two key differences – choice of baseline and methodology – could be expected to have the following effects:
In any given week these two offsetting factors could result in the ONS measure of excess deaths being higher, lower or the same as the CMI’s measure.
The ONS approach does not answer the same question as the CMI approach. The ONS approach is not trying to tell us how current deaths compare to a pre-pandemic year. It does tell us whether deaths are higher than some recent history. Using a rolling five-year average is not an uncommon approach to calculating demographic assumptions and helps to smooth out some of the highs and lows of individual years. As long as you trust that the five years chosen for the average contain information about how current deaths might be expected to develop, this can be a useful approach.
The Department of Health, via the Office for Health Improvement and Disparities (OHID), also produces excess death statistics. The OHID information contains useful breakdowns of deaths’ data, for example by cause of death, deprivation or region. However, the OHID definition of an excess death is different again to the CMI or ONS approaches. OHID produced a model to estimate expected deaths using five years’ worth of data from 2015 to 2019.
The longevity landscape is currently very complex, with many pressures acting on mortality rates. Factors ranging from NHS pressures and cost of living to extreme weather events, as well as ongoing direct deaths from the pandemic, can all contribute to the deaths occurring in a year.
We shouldn’t necessarily expect one single measure to give the complete picture of everything going on currently. The most appropriate measure will depend on exactly what question we are seeking to answer. Indeed, there is an ongoing collaboration led by the ONS with CMI involvement to review the various methods available to deliver improved insight on excess deaths.
The Office for Statistics Regulation also recently reviewed some of the excess mortality data available.
