Claim your CPD points
In summary:
|
The COVID-19 Mortality Working Group has examined the latest Provisional Mortality Statistics, covering deaths occurring prior to 30 November 2022 and registered by 31 January 2023, released by the Australian Bureau of Statistics (ABS) on 24 February.
This release also included the article COVID-19 Mortality in Australia, with details on all COVID-19 deaths occurring and registered by 31 January 2022.
This is our first detailed analysis since the 31 August data was released. Therefore, much of our commentary focuses on the three months to 30 November 2022. We also provide an indicative estimate of excess mortality for the month of December 2022.
Consistent with the terminology used by the ABS and reflective of the data provided, throughout this article, we separate COVID-19 deaths into:
In this article, we calculate excess deaths by comparing observed deaths to our “baseline” predicted number of deaths for doctor-certified deaths (by cause) and coroner-referred deaths (for all causes combined).
The Working Group has also examined excess mortality by age/gender and by state/territory. In so doing, we have used additional data supplied by the ABS in a customised report in relation to COVID-19 deaths registered by 31 January 2023, namely:
Our previous Actuaries Digital article discussed in some detail how we arrived at our baseline predicted deaths. In short:
The baselines for our estimates of excess deaths remain “in the absence of the pandemic” for each of the three years 2020 to 2022. We have not included any COVID-19 deaths in the baseline, as these would not exist in the absence of the pandemic.
As always, it is important to note that predicted death numbers are increasing faster from demographic changes (ageing and population size) than they are reducing due to mortality improvement. Therefore, our model predicts higher baseline numbers of deaths in each successive year.
The results we present here differ from those quoted by the ABS in its commentary surrounding the release of the Provisional Mortality Statistics. The ABS explicitly states that the comparison presented is not an official estimate of excess mortality and that their baseline for 2022 (being the simple average of the number of deaths from 2017, 2018, 2019 and 2021, with no allowance for mortality trends or demographic changes) serves as a proxy for the expected number of deaths only.
For example, the latest Provisional Mortality Statistics says: “In 2022, there were 174,717 deaths that occurred by 30 November and were registered by 31 January 2023, which is 22,886 (15.1%) more than the historical average.” By comparison, we estimate that there were 17,900 (11.3%) excess deaths in this period. The main reason for the difference between the ABS figure of 22,886 and our estimate of excess deaths of 17,900 is the difference in the baselines used.
Figure 1 shows the results of our analysis, comparing actual deaths each week to our predicted values and the 95% prediction interval.
Deaths in September and October, while higher than the predicted level, were within the 95% prediction interval for all but one week. This is in stark contrast to earlier in the year when almost every week was well above the prediction interval (i.e. above the 97.5thpercentile). In November, however, the last three weeks were again well above the upper end of the prediction interval.
Figure 2 shows deaths from COVID-19 and with COVID-19, noting that, given small numbers, weekly data for with COVID-19 deaths is only available for 2022.
* COVID-19 data from ABS customised report 2023
Deaths from COVID-19 peaked in the last week of July, trended downwards until the end of October, and then rose again throughout November. Deaths with COVID-19 have followed the same pattern, although with a little more volatility given the smaller numbers of such deaths.
It is unclear how many people who died with COVID-19 would have died during this period anyway, and how many may have had their death hastened by COVID-19. However, given that with COVID-19 deaths follow the same peaks and troughs as from COVID-19 deaths, and that with COVID-19 deaths are only counted as such if COVID-19 is considered to have contributed to death, it seems that COVID-19 is a catalyst for with COVID-19 deaths, rather than being merely coincidental. We note that the same traditionally happens with the winter peak of respiratory disease deaths.
There may be instances where it is difficult to determine whether a death is due to COVID-19 versus COVID-19 being a contributing cause, and an element of professional judgment by the certifier will necessarily need to be exercised. It is unclear to us how often such decisions may need to be made, and how much “blurring” there may be between from COVID-19 and with COVID-19 deaths.
Figure 3 shows the comparison of actual deaths to predicted after removing from and with COVID-19 deaths.
Even after removing all from and with COVID-19 deaths, significant excess mortality remains for 2022, with:
* COVID-19 data from ABS customised report 2023
This can be seen clearly in Figure 4, where we divide excess deaths into those from COVID-19, those with COVID-19 and the residual excess from other causes, set against the 95% prediction interval. The correlation of the trend of non-COVID-19 excess deaths with COVID-19 deaths can be seen in the 9-week centred rolling average. The next section discusses the causes of death driving this trend.
Table 1 shows our estimate of excess deaths broken down by cause.
As in our previous work, we have assumed that coroner-referred COVID-19 deaths will be 5% of all COVID-19 deaths in 2022, based on the experience of late 2021 and the emerging experience in 2022. If our estimate of coroner-referred COVID-19 deaths is too high (or low), this will not affect the total level of excess deaths measured; it will just mean that our estimate of non-COVID-19 coroner-referred deaths will be too low (or high) by the same amount.
Figure 5 and Figure 6 show weekly excess deaths by underlying cause (thus including deaths with COVID-19) since the start of the pandemic, compared with the 95% prediction interval. All graphs are shown using the same y-axis to give a sense of the contribution of each cause. The 9-week centred average is also shown, to highlight any trends.
Figure 5 shows a clear increase in excess deaths due to ischaemic heart disease from around March 2021 until around August 2022, with the peaks and troughs following the same timing as deaths from COVID-19 (noting that each chart includes deaths from that cause with COVID-19). There has been a consistent, albeit volatile, excess for deaths from both cancer and cerebrovascular disease since around July 2021. Diabetes deaths have generally been higher than expected throughout the pandemic, and again the peaks and troughs have followed the same patterns as deaths from COVID-19.
Of these causes, ischaemic heart disease is the biggest contributor to excess deaths in 2022, followed by cancer.
Figure 6 shows negative excess deaths for dementia in 2020, 2021 and September/October 2022 (closely correlated with lower respiratory disease), with higher excess deaths in the first half of 2022 (closely correlated with COVID-19 and flu waves). Deaths from respiratory disease have been significantly lower than expected throughout the pandemic, except for the short (and early) flu season which peaked in late June 2022. This coincides with the second 2022 peak in excess non-COVID-19 deaths that we saw in Figure 3and Figure 4.
Deaths from other diseases (where available ABS data does not specify the cause) were also lower than predicted in 2020 (correlated with lower respiratory disease) but have been, as a group, the largest contributor to non-COVID-19 excess deaths in 2021 and 2022. It is not clear what might be driving this, although we expect that at least part of the excess will be in respect of people who otherwise may have succumbed to respiratory disease in 2020 and 2021. We also note that deaths from heart conditions other than ischaemic heart disease traditionally comprise more than 25% of this group and that there does appear to be an increased risk of heart disease after recovering from COVID-19.
Coroner-referred deaths include deaths from COVID-19. In Figure 6, we have separately shown these (in red), using actual coroner-referred COVID-19 deaths for 2020 and 2021, and estimates for 2022 based on our assumption that 5% of all COVID-19 deaths are referred to the coroner. The residual non-COVID-19 coroner-referred deaths are well above expected levels in the second half of 2021 and throughout 2022. We note, however, that suicide monitoring reports for NSW and Victoria show a slight increase in 2022, but not large enough to account for the level of excess seen in those states. Similarly, Australia-wide road death statistics show a small increase in 2022, but not large enough to account for the overall increase.
While the ABS Provisional Mortality Statistics data is only available up to the end of November 2022, we have formed an estimate of excess deaths in the month of December 2022 to arrive at a full-year estimate for 2022.
We have:
This estimation process results in total excess deaths for the month of December of 14% of predicted, half of which (7% of predicted) are from COVID-19 and another 15% (2% of predicted) are with COVID-19.
For the whole of 2022, our estimate is that excess deaths were 12%. In other words, there were nearly 20,000 more deaths in 2022 than would have been expected if the pandemic had not happened. Of these, 10,300 were from COVID-19, with another 2,900 with COVID-19. The remaining 6,600 excess deaths were from other causes, with 1,500 fewer deaths than expected from respiratory causes and 8,100 excess deaths from other non-COVID-19 causes.
* COVID-19 data from ABS customised report 2023
Table 2 below shows our estimate of excess deaths by age band and gender. We have shown the excess including all deaths, and then again after deducting from and with COVID-19 deaths.
* COVID-19 data from ABS customised report 2023
Almost all age bands show excess deaths in 2022 (which is not all that surprising given the excess deaths for the whole population), but the number and percentage of excess deaths is higher in older age bands. The risk of mortality from COVID-19 is steeper than the underlying age mortality curve, so this result is somewhat expected.
However, it is notable that there are excess deaths in all age groups and that this excess is generally significant, even after removing COVID-19 deaths. To look at this further, we have shown these results similarly to the graphs by cause, showing weekly excess deaths by age band since the start of the pandemic. For these charts, we have been able to show the contribution of deaths from and with COVID-19 (in orange) separately to non-COVID-19 deaths. The 95% prediction interval is also shown. All age bands are shown using the same y-axis to give a sense of contribution of each age band.
* COVID-19 data from ABS customised report 2023
Figure 8 shows that excess deaths in 2022 are dominated by the older age groups. However, we already expect many more deaths in these age groups. It is instructive to consider the same information with the excess deaths expressed as a percentage of predicted deaths.
* COVID-19 data from ABS customised report 2023
Figure 9 shows that the older age groups still experienced the most significant increase in excess deaths when expressed as a percentage of predicted deaths. Indeed, deaths of people over the age of 75 are significantly higher than expected in almost every week of 2022.
While the numbers of deaths in the 0-44 and 45-64 age bands are small, we saw in Table 2 that year-to-date excess deaths in 2022 are materially higher than expected. That table also showed that the percentage excess was higher for females than for males. The differences are worth investigation, although the small numbers mean that there is considerable natural variation.
* COVID-19 data from ABS customised report 2023
Figure 10 shows that female non-COVID-19 mortality experience in 2021 and 2022 is noticeably worse than male, especially in the 0-44 age band.
Table 3 shows our estimate of excess deaths by state/territory, before and after deducting deaths from and with COVID-19. Data is grouped for smaller states/territories, where small numbers meant that the ABS could not give us detailed COVID-19 figures.
* COVID-19 data from ABS customised report 2023
In 2022, all states/territories apart from NT had significant levels of excess mortality ranging from 10% to 15% of predicted. Generally, about half of this is due to deaths from COVID-19, with another 1-2% due to deaths with COVID-19.
The graphs below show these results week-by-week (with the excess shown as a percentage of the predicted value) for:
* COVID-19 data from ABS customised report 2023
Figure 11 shows that these larger states all had better-than-expected mortality in 2020. It also shows the impact of the second COVID-19 wave in Victoria.
In 2021, Victoria shows a much higher level of excess deaths in the Delta wave – in the last quarter of the year – than either NSW or Queensland. A large portion of these deaths do not have COVID-19 on the death certificate.
In 2022:
* COVID-19 data from ABS customised report 2023
The SA experience is similar to the three largest states, although somewhat more volatile given its smaller size.
Western Australia, while mostly having excess deaths within the 95% prediction interval, had many more weeks with positive excess deaths than with negative excess deaths. There is no large COVID-19 spike in January/February, thanks to the later opening of WA’s borders.
Figure 13 shows the higher volatility of excess mortality experienced by the smallest state and the two territories. Allowing for that volatility, Tasmania had broadly the same experience of excess mortality as the larger states, but with a high level of excess deaths during the Delta wave late in 2021.
With its relatively young and affluent population, the ACT has experienced lower excess mortality than the larger states.
The Northern Territory has a very young population, which might explain the low net impact of the pandemic after allowing for the high volatility caused by a very small weekly expected death count. There is no apparent general trend in excess mortality, although it does appear to have been declining since the middle of 2022.
The measurement of higher numbers of deaths than predicted does not tell us why this is occurring. There are several reasons hypothesised around the world (where this effect is occurring to a greater or lesser extent). It isn’t possible to identify from death counts alone what is causing the non-COVID-19 excess deaths, but we have listed below the most likely explanations. We note that multiple factors are likely in play, and different factors may be more or less pronounced at various times.
To summarise the evidence from earlier sections:
Based on these observations, the following indicates which factors, in our view, are likely to be having a greater or lesser impact on Australian excess mortality in 2022.
Figure 14, which shows excess deaths from doctor-certified respiratory disease, other non-COVID causes and those from COVID-19, puts the above into context, across the pandemic. Note that deaths with COVID-19 are included with the primary non-COVID cause (respiratory or other).
* COVID-19 data from ABS customised report 2023
Figure 15 shows the cumulative standardised mortality rates (SDRs) for 2015 to 2022, expressed relative to the rate for 2019. The SDRs are from the Provisional Mortality Statistics, plus allowance for late-reported deaths.
The graph shows that:
One question that regularly crops up is why COVID-19 deaths reported by Victorian health authorities appear to be so much higher than those reported by NSW. The customised data that we have received from the ABS allows us to shed a little light on this, although some questions still remain.
The customised data that we have received from the ABS includes deaths both from and with COVID-19 for each of NSW and Victoria for 2022, with the deaths summarised into the week that they occurred.
From the Federal Department of Health website we also have 7-day averages of the surveillance deaths reported by each state/territory health authority, summarised by the date the death occurred. The 7-day averages are reported in whole numbers only, but by multiplying by 7 we get an approximate number of surveillance deaths reported for each week of occurrence.
Per the national surveillance definition, deaths are reported as COVID-19 deaths if a person died with COVID-19, not necessarily because COVID-19 was the cause of death. Deaths are only excluded if there is a clear alternative cause of death that is unrelated to COVID-19 (e.g. trauma). As such, surveillance deaths will include deaths from COVID-19, deaths with COVID-19 and other deaths where the doctor/coroner has determined that COVID-19 was incidental and had no role in the death of the person.
By comparing the two datasets, we can estimate the number of surveillance deaths reported that do not end up with COVID-19 recorded on the death certificate.
Figure 16 shows, for each of NSW and Victoria, surveillance deaths expressed as a percentage of predicted deaths, broken down into those from COVID-19, those with COVID-19 and those where COVID-19 does not appear on the death certificate.
* From and With COVID-19 data from ABS customised report 2023
We can see that:
We do not have an explanation for why Victoria would record more surveillance deaths than NSW.
This monthly COVID-19 mortality analysis is intended for discussion purposes only and does not constitute consulting advice on which to base decisions. We are not medical professionals, public health specialists or epidemiologists.
To the extent permitted by law, all users of the monthly analysis hereby release and indemnify The Institute of Actuaries of Australia and associated parties from all present and future liabilities that may arise in connection with this monthly analysis, its publication or any communication, discussion or work relating to or derived from the contents of this monthly analysis.
The members of the Working Group are: