Part 3 - A Nation Grows Older: Disparate Mortality Trends and Other Demographic Curiosities
- an Irishman dons his hobby statistician's cap (again)
If you have been following this series you will recall we have identified multiple simultaneous trends in official Irish population and mortality data:
Overall population growth ↗
Significant growth of elderly cohort ↗
Increasing annual deaths totals ↗
Decreasing annual percentage death rates ↘
In addition:
2020 deaths data shows no indications of a once-in-a-generation pandemic ¯\_(ツ)_/¯
This series of posts attempts to to better understand if there is excess mortality in the Republic of Ireland, to quantify it, and to locate it in the Irish population structure. See Part 1 here, and Part 2 here:
Continued…
Trend Lines: Which are the ‘expected’ mortality rates?
Consider the plots of % mortality rates again, this time with trend lines added for 2011-2020 and extended to 2021-2022:
Remarkably, the extrapolated 10-year linear trend line shows declining mortality rates across all age groups - this points to improving health and life expectancy across the whole Irish population (note, these are simple auto-generated linear trend lines which seem to fit the age groups 40 years and older very well.) The actual mortality rates oscillate a little above and below the trend line, sometimes mortality rates are above trend and then below again, or remain temporarily stable in-between.
Now, just as we criticised Eurostat and Euromo in Part 1 for ignoring an upward trend in Irish deaths totals resulting in overestimated excess deaths, if we ignore the downward trend in percentage death rates per age group, we could be guilty of underestimating excess deaths. Applying our simple linear trend lines we can model expected mortality rates for 2021 and 2022 using mortality rates extrapolated from the trend lines.
Proceed with Caution!
Trend lines of past data are not necessarily predictive, and there is no rule by which mortality rates should keep falling (or rising) just because they did so in prior years. Furthermore, it is obvious percentage mortality rates cannot fall indefinitely and in many age cohorts the decline has been quite slight, so a levelling off cannot automatically be considered unusual. That said, a simultaneous change of trend across many age groups would be significant as would any reversal of trend (declining death rates flipping to increasing death rates). With the caveat that 2021 and 2022 figures are still provisional, we can now make a provisional analysis.
Modelled Expected Deaths and Excess Deaths
Applying the 2011-2020 linear trend line values for % death rates per age group results in the following (red indicates excess deaths, green deficits). Note: 2021 and 2022 official deaths totals are incomplete and still provisional.
Although 2021 and 2022 death figures remain provisional, most age groups have already exceeded our modelled expected deaths and therefore may have been experiencing excess deaths. Remember, this analysis is very dependent on the assumption that death rates were expected to continue to fall.
Alternatively, if we assume mortality rates were levelling off in 2020, and we apply the trend line percentage death rates from 2020 to the 2021 & 2022 populations we get the following:
As mentioned, the 2021 and 2022 totals are all subject to upward revisions. This means the rows for individual age cohorts represent a best case scenario (every extra death added to an age group will shift the calculation in the direction of more excess deaths). The final rows represents the worst case scenario at the total population level. [Edit: And the second to last row “All ages Total” represents excess deaths with the current provisional totals.]
Late Registered Deaths
The uncertainty about the deaths figures for 2021-2022 arises out of a peculiarity in the Irish statistical reporting of deaths whereby deaths awaiting a coroner’s certificate cannot officially be registered. Thus, those deaths awaiting a coroner’s certificate are not included in the official statistics - this process can take years and has been affecting an increasing percentage of death certificates recently. From the General Register’s Annual Report 2021:
The Central Statistics Office is responsible for the publication of official mortality statistics for Ireland. As deaths can be registered up to three months after the date of death, and as registration of a death may be delayed where the death has been referred to a Coroner for investigation or inquiry, the CSO publishes quarterly deaths approximately five months after the end of a reference quarter. These initial quarterly publications are based upon registration data. The CSO subsequently publishes annual statistics based on deaths by date of occurrence, approximately 22 months after the end of a reference year.
The General Register’s Office neglects to mention here that the CSO also updates the figures released in the Vital Statistics Annual Reports on an ongoing basis, for example in the dataset VSA35 “Revised Deaths Occurring” (used in these substack posts), and in the Appendices to subsequent Vital Statistics Annual Reports. Judging by recent years, over 99% of all deaths are normally registered by the end of the [EDIT: second third] calendar year following the death - e.g. by the end of 2022 the deaths totals for 2019 should be over 99% complete.
==UPDATE 11.12.2023== Frustratingly, we have had to revise down our estimation of how complete the 2020 data is. In previous years, 34 months after calendar year’s end, officially released figures were always over 99% complete. However, based on the CSO’s own analysis of the cleaned RIP.ie data, the 2020 figures are almost certainly still less than 99% complete and possibly still less than 98% complete. ==
Considering the following, it is not surprising deaths requiring a coroners’s certificate have been rising [emphasis by Substack authors]:
Deaths are required by law to be referred to a coroner where the death has occurred in suspicious circumstances, was unexpected, unexplained, or the deceased was a resident of a nursing home or died of a hospital acquired infection. Covid-19 related deaths or where death was suspected of having been caused by Covid-19 required to be reported to Coroners.
The number of deaths registered on foot of a coroner’s certificate has risen significantly over the years and represented near 40% of all deaths registered during 2021 – Table 27 – in part reflecting the requirement to refer deaths to a Coroner where Covid-19 was suspected as being a contributory factor.
Furthermore, late registered deaths from previous years demonstrate how, due to the small number of deaths in the younger age groups, relatively small additions could easily tip those cohorts into excess deaths also. Indeed, the younger cohorts are disproportionately subject to more late registered deaths than the older cohorts. For example, applying the proportions of late registered deaths per age group in 2019 to the pending deaths for 2021 suggests it possible almost all age cohorts be shaded red in the above tables indicating excess deaths once the late registered deaths are added.
Online Deaths Notices
RIP.ie is an Irish obituary website dedicated to publishing death notices in Ireland. The website allows funeral directors to post death notices on the website without additional costs to the family. As such, its death notices offer a very good proxy for estimating mortality trends. From the Irish Central Statistics Office (CSO) report, “Measuring Mortality Using Public Data Sources 2019-2023 (October 2019 - June 2023)” published in October 2023:
Due to the Irish custom of holding funerals within two to three days following a death, these notices are usually placed in a fast and efficient manner, providing a valuable crowd-sourced means of tracking deaths. The notices are placed close to 'real time'. We found that the average length of time between date of death and publication of the notice on the website is about 1.3 days. This most recent analysis shows that death notices provide accurate estimates up to 12 months earlier than official data.
Crowd-Surfing Irish Mortality Trends
The website IrelandExcessDeaths.com claims to have cleaned - removing any duplicates or deaths outside of the Republic of Ireland - from the RIP.ie publicly available deaths notices and reports deaths totals for 2021 and 2022 well in excess of the official data from the GRO and CSO. Reviewing the figures claimed for 2015-2020 which matched closely (slightly less than) the official figures, and considering the CSO’s report conclusions, suggests the figures from IrelandExcessDeaths.com may be reliable. It is from their figures we have approximated the pending death registrations for 2021 and 2022 at ~1300 and ~1800 respectively.
Note: This Substack’s authors do not endorse IrelandExcessDeaths’ conclusions as they are based solely upon raw deaths totals, using 5-year pre-pandemic averages as a baseline and fail to adjust for population growth, population structure, or mortality rate trends.
Dry Tinder Effect
An additional phenomenon worth mentioning regarding mortality trends is that the expected normal number of deaths will rarely be 100% accurate, it will nearly always be too high or too low in comparison to actual deaths and unexpected events can/will cause larger deviations. Accordingly, that rates in 2021 may rise above the trend line is not entirely unexpected in some age groups. Epidemiologists sometimes refer to this as the “dry tinder effect” meaning a year of less than usual mortality leads to a larger group of susceptible individuals the following year. The metaphor arises from the observation that larger forest fires in a given year can often be explained by fewer fires in previous years.
Of course, the converse can also be true, where a year of excess deaths is followed by a return to the trend line or sinking below it, constituting a year of deficit deaths (i.e. less than expected).
Conclusions (Provisional)
We have tried to better understand if there is excess mortality at all, to quantify it, and to locate it in the Irish population structure. Exploring the extent of the rapid population growth in 21st century Ireland, seeing the massive changes in the population structure and also realising the long-term natural aging of the population starting to manifest itself in annual deaths has been an eye-opener.
Some of this analysis may have been counter-intuitive and any criticisms or suggestions are very welcome. As 2021-2022 deaths figures remain incomplete and provisional, so too must these conclusions remain tentatively provisional.
Firstly, in this assessment there are no signs of an alarmingly deadly pandemic in 2020 in the deaths data for Ireland. In contrast, Ireland experienced comparatively elevated percentage death rates in 2011-2013 but there were no pompous talking heads blaring their panic nor odious dashboards blinking death totals back then.
There is a legitimate discussion to be had about which mortality rates are representative of a “normal” mortality year in Ireland during the pandemic 2020-2022. Sample scenarios:
- Assuming mortality rates were expected to continue to decline results in notable levels of excess deaths for 2021 & 2022 (presumably across all age groups, once late registered deaths are included).Assuming mortality rates were expected to level off (somewhat unrealistic in the younger age groups under 40, more plausible in the older cohorts over 40) still results in excess deaths, though to a lesser extent.
Note: the mortality rates in 2020-2022 are not in and of themselves remarkable (e.g. compare with 2011-2013 levels), it is the change in trend which is significant. This observation was regrettably missed in our initial analysis (now unpublished but updated and being reposted here in Parts 2 and 3.)
This series has not (yet) looked at deaths in 2023. The CSO has this to say in October 2023, “A pronounced increase in the number of death notices was observed in late December 2022 and early January 2023.”
Further Posts?
Some further aspects of mortality trends that have not been reviewed in this series are:
Temporal Analysis of deaths on a monthly or weekly basis to identify when the excess deaths have been occurring. It would be particularly interesting to see when the excess deaths started in the various age groups.
Mortality Years (July-June) instead of calendar years (January-December) avoid bridging winter seasons (when mortality naturally increases) and allow seasonal deviations to even out.
Life Expectancy and lost life years which offer an alternative method of measuring the extent of mortality trends by weighting younger deaths more when assessing overall mortality.
Finally, we take no pleasure in pointing this out, and to repeat what many in the Irish Dáil (parliament) Debate in October of this year stated: overall death totals in Ireland will likely continue to worsen (i.e. increase) as the population continues to age, with cohorts moving up the population pyramid into the elderly age groups where the vast majority of deaths occur.
Please share this series widely if you feel it helpful in exploring and explaining the phenomenon of mortality trends in the Republic of Ireland. We hope to update this series as new data becomes available.
Nice work, Witzbold! I would love to hear any feedback on my own attempt to make sense of Canada's (much less detailed) data:
https://thoughtsonacrisis.substack.com/p/mortality-mismatches-in-canada
A question about trendlines from a fellow amateur: would it not be the case that any population shifts over time that are happening at a more or less regular rate (even a regularly *increasing* rate) end up being 'baked into' raw mortality figures? I understand the concept of a shifting population pyramid and an aging population, but unless there are some external factors (major shifts in immigration; war; etc.) wouldn't a trendline from a reasonably long period predict the immediate future fairly well? It seems to me that for the purposes of assessing excess mortality during what I call the 'Covid-response' period (starting in March of 2020) an extrapolated trendline should be enough to pick up major deviations. But this is all just my sense of how things should work, based on zero training or background in demographics! I'd love to hear your take on it, and on the Canada situation as I've described it.
Very thorough analysis, thank you!
Now do births for Ireland... :)
But seriously: numbers of births are one thing, but birth rates (number of births divided by number of potential mothers, maybe even considering age of mother at birth) are more informative. After a short baby-boom 2021/22, births in Ireland now seem on the way down as well, but maybe this can be explained by population structure.