Measuring Australia's excess mortality during the COVID-19 pandemic until the first quarter 2023

Weekly and annual excess mortality estimates for Australia and states during the COVID-19 pandemic until the first quarter of 2023

Cite
Print
Released
19/07/2023
Source
Provisional Mortality Statistics, Jan - Mar 2023

Key Statistics

Overview

The COVID-19 pandemic has impacted the lives of people in Australia. Excess mortality, typically defined as the difference between the total number of deaths in a specified period and the expected numbers of deaths in that same period, continues to provide important insights into the mortality profile in Australia during this time. Since the start of the Omicron period in Australia (Jan 2022-current), excess mortality has been recorded for all jurisdictions in Australia. The map below shows the percentage of excess mortality above the expected number of deaths during the Omicron period across states and territories in Australia. The map is designed to illustrate the differences in excess mortality across jurisdictions - it does not represent statistically significant differences.    

This article presents weekly and annual excess mortality estimates for Australia during the COVID-19 pandemic until the first quarter of 2023 for Australia and states and territories. For the purposes of this article, the first quarter of 2023 refers to the first 12 weeks of 2023, up until 26 March 2023. Expected mortality in this article is modelled as being in the absence of the COVID-19 pandemic.

Key points:

  • Excess mortality has continued for the first quarter of 2023.
  • Excess mortality in 2023 is lower for most jurisdictions, compared to the same period in 2022. 
  • COVID-19 is still a significant contributor to increased mortality. 

 

Excess mortality (%) by jurisdiction, Omicron period (Jan 2022 - Mar 2023)

This map displays the percentage of excess mortality in each state and territory in Australia in 2022 and 2023. This time frame corresponds with the Omicron period in states and territories in Australia.

COVID-19 pandemic in Australia

Australia has had multiple waves of COVID-19 infections since the start of the pandemic in March 2020.

The first wave was recorded from mid-March to mid-April, with most states in Australia recording active infections and associated mortality. The predominant variant during Wave 1 was the original virus strain. Nationally, public health measures were implemented which included social distancing and hygiene recommendations, stay at home measures and shut down of national borders. 

The second wave started in June 2020, with the largest number of active infections and deaths occurring in Victoria. The variant during Wave 2 was the original virus strain. ‘Stay at Home’ orders were implemented in Victoria from July 2020 until the end of October 2020. While stay at home measures were lifted across other jurisdictions, there were still strict travel restrictions in place nationally. 

Between the second and third wave, many jurisdictions managed smaller outbreaks with localised lockdowns.

A third wave began in June 2021 with the spread of the Delta variant. Most jurisdictions recorded COVID-19 infections during this period. New South Wales and Victoria recorded the highest number of infections and mortality due to COVID-19 during the Delta variant wave and both states implemented public health measures to manage this, including vaccinations, social distancing and stay at home measures. 

In November 2021, a staged re-opening of the international borders began. The Western Australia border was the last to re-open on 3 March 2022. The re-opening of borders coincided with high vaccination rates in Australia. 

The Omicron variant was first identified in Australia in November 2021 and began to spread throughout the country at the end of 2021, continuing into the first quarter of 2023. There have been multiple waves associated with the Omicron variant with waves often being associated with a new strain. 

Key events during the COVID-19 pandemic in Australia are noted in graphs where relevant.

The following table shows excess mortality for each jurisdiction expressed as the percentage above expected for that year.

Excess mortality as a percentage above expected by jurisdiction, 2020-23
 2020202120222023
Australia-3.11.410.99.1
New South Wales-4.10.110.78.6
Victoria-0.93.413.212.0
Queensland-4.30.810.17.8
South Australia-3.20.59.28.9
Western Australia-3.90.66.26.1
Tasmania-3.65.813.617.3
Northern Territory1.56.810.6np
Australian Capital Territory-4.3-2.812.18.9

a. Data is provisional and subject to change. 
b. Years are based on a sum of ISO weeks derived from the weekly modelling. There are 53 weeks in 2020. There are 52 weeks in 2021 and 2022. Excess mortality has been estimated for the first 12 weeks of 2023. 
c. Deaths in 2023 are deaths that occurred by 26 March and were registered and received by the ABS by 31 May 2023.

Introduction

Excess mortality measures have been a common statistical tool used around the world to understand the full impact of the COVID-19 pandemic on mortality. Excess mortality is defined as the difference between the total number of deaths from all causes in a specified period and the expected numbers of deaths from all causes in that same period. Excess mortality measures can account for deaths due to COVID-19, potentially misclassified or undiagnosed COVID-19 deaths, and other mortality that may be indirectly related to the pandemic (e.g. relating to social isolation or changed access to health care). For these reasons, excess mortality is generally considered a more comprehensive measure for assessing the impact of the pandemic rather than considering only the number of deaths due to COVID-19. 

As the pandemic progresses, the use of excess mortality has become even more pertinent as a measure. Attributing COVID-19 as a direct or contributing cause of death is dependent on a number of factors including rates and types of testing (e.g. Rapid Antigen Test versus Polymerase Chain Reaction testing), diagnostic ability (e.g. having the ability and resources to diagnose the train of morbid conditions leading to death), and infection prevalence. Even in Australia, a country with a strong health care system and robust certification of deaths, the attribution of COVID-19 to a death can be difficult to assess. Reasons for this include a patient having multiple co-morbidities, doctors being unaware of a previous COVID-19 infection or difficulties in assessing the impact of an acute infection compared to a chronic infection (e.g. long COVID). Changes in the attribution of COVID-19 to death are evident in the data. In 2023, 73.7% of COVID-19 associated deaths were "from COVID-19", where the deceased has been certified as dying as a direct consequence of the virus. In 2020 this percentage was 99.0%. There have also been increases in the number of incidental mentions of COVID-19 as part of the death certification process. These deaths (coded to ICD-10 code U08.9, Personal history of COVID-19) are not included in official COVID-19 reporting. See article, COVID-19 Mortality in Australia for more information. 

Taking the above into account, this article provides excess mortality estimates for Australia and states and territories until the first quarter 2023. Information on how to interpret and use the excess mortality estimates is also provided. The ABS will produce another excess mortality report in the second half of the year. This second report will include additional analysis on diseases and demographics such as age groups. 

Measuring excess mortality

Across the world, health and statistical authorities have sought to measure excess mortality during the COVID-19 pandemic. Different methodologies have been applied, with the goal to predict an expected number of deaths for a given year. Choice of model and baseline can markedly affect estimates of expected (and therefore excess) mortality. Estimating the expected future seasonality for deaths can be a challenge in many models. The suitability of a model can depend on factors such as country context, data quality and collection methods.

The ABS has adopted aspects of a methodology used by New South Wales (NSW) Health, applying a cyclical linear regression with a robust estimation procedure to produce both an expected number of deaths and a range of expected deaths to the first quarter 2023. This model was also used for four previous Excess Mortality reports released by the ABS. The ABS has updated the model this time. Changes include: 

  • All-cause mortality includes deaths certified by both a doctor and a coroner. Previous reports analysed doctor certified deaths only. 
  • Age-specific death rates (ASDR) are used to model counts. Each ASDR is modelled separately to obtain an expected rate of death for that age group. Rates are reverted to counts of deaths and then added up across age groups to obtain an expected count for the total population. Previously, counts of deaths were modelled only. Using ASDRs allows for changes in population size and age composition. Further information is provided in the Methodology section. 
  • The baseline period used to predict the expected number of deaths in 2020-2023 includes the years 2013-2019. Previously, the baseline period started at either 2015 or 2016. 

See the methodology section at the end of this article for more information. 

Choice of mortality measure

Selecting the right mortality measure to answer your research question is important. Excess mortality measures are just one way to gain an understanding of what may be occurring in a population and how this has changed over time.

Other mortality frequency measures that may be used to assess changes in patterns of mortality may include age-standardised rates (SDRs), age-specific rates (ASDRs), years of life lost and life expectancy. Both ASDRs and SDRs are presented per 100,000 persons in this article.

A mortality rate is a measure of the frequency of occurrence of death in a defined population during a specified interval. Mortality rates take into account different population sizes and enable comparisons across cohorts and over time. Mortality rates can include crude rates, age-specific rates, cause specific rates and age-standardised rates. Analysis of mortality is strongest when mortality frequency measures are used in combination. Updated causes of death for 2022, as well as analysis of years of life lost and official life expectancy estimates will be released in the last quarter of 2023.

ASDRs are rates calculated by dividing the number of deaths occurring in a specified age group (e.g. 85-89 years) by the population in that age group. ASDRS are useful for understanding the mortality pattern in a specific age group. For causes of death which affect a particular age cohort most commonly (e.g. dementia), ASDRs can be a useful measure to assess burden.

SDRs enable comparison of populations with different age structures. They are a weighted average of ASDRs where the weights are taken from a standard population. The ABS uses the population in 2001 to standardise age-specific rates and produce SDRs. SDRs are an extremely important frequency measure of mortality. Not only do they enable comparison between cohorts and over time, they allow analysis of trends over time. Australia has a low mortality rate and over time the general trend has been for the SDR to decrease. In 2022, there was an increase in Australia's SDR.  

See the table below for the SDRs over the pandemic period from all causes of death and from COVID-19. See the methodology section for a 10 year time series of SDRs by state and territory.

Number of deaths and standardised death rates by jurisdiction, 2020-22
 Number of deaths  SDRs  
 202020212022202020212022
All deaths      
Australia162,675172,096191,049495.6508.9547.7
New South Wales52,88855,67862,284489.9501.7546.3
Victoria41,03243,07247,557484.9498.3536.2
Queensland31,93534,35838,544510.8529.0568.5
South Australia13,78914,42615,935510.5519.3556.9
Western Australia15,08216,09317,422471.1481.6502.7
Tasmania4,4084,8025,141534.3562.9584.1
Northern Territory1,1421,2061,281695.8717.1737.4
Australian Capital Territory2,3992,4612,885524.1519.8585.6
Deaths from or with COVID-19      
Australia9151,41513,1542.64.336.3
New South Wales646534,7900.66.240.2
Victoria8137413,4429.08.737.1
Queensland432,321npnp33.3
South Australia441,044npnp34.4
Western Australia110931np—26.1
Tasmania170335np—38.0
Northern Territory0185—np57.9
Australian Capital Territory213206npnp41.0

a. Data is provisional and subject to change.
b. Years are calendar years, not based on ISO weeks.
c. Deaths from COVID-19 have been coded to ICD-10 code U07.1, U07.2 or U10.9. Deaths with COVID-19 have an associated cause of any of ICD-10 codes U07.1, U07.2 or U09.
d. Includes all deaths that occurred and were registered by 31 May 2023.

Interpreting results

What are we measuring?

Having a clear research question and understanding what we are measuring is imperative for selecting the right statistical tools. For excess mortality, the research question will also influence what inputs go into the model, for example, the baseline years used to predict the expected number of deaths. 

The most common research question being assessed with excess mortality analysis has been: 'How does the number of deaths which has occurred during the COVID-19 pandemic (2020-2023) compare to the number of deaths expected had the pandemic not occurred?'. For this research, expected deaths are modelled using data from before the pandemic (usually ending at 2019). This information can then be used to assess the impact of a multitude of factors during the pandemic, including stay at home measures, changes in virus strain and other public health responses. 

Recently, there is a second research question of interest: 'If COVID-19 is now part of our mortality profile, should a new expected number of deaths be calculated, which incorporates deaths from the virus into the count?'. For this research, expected deaths would need to be modelled including COVID-19 in the baseline. This information could be used for planning (e.g. how many deaths are expected while COVID-19 is circulating) and give an indication of what is happening currently. Importantly, this information could not be used to answer the first research question regarding expected deaths in the absence of a pandemic because the baseline in this model would include COVID-19 deaths. 

The ABS considered both research questions. It was decided that the focus of this paper will remain as 'How does the number of deaths which has occurred during the COVID-19 pandemic (2020-2023) compare to the number of deaths expected had the pandemic not occurred?'. This is for several reasons:

  1. While there have been fewer deaths in 2023 from COVID-19 compared to 2022, the virus is still having considerable impact on the mortality profile in Australia.
  2. Some of the broader impacts from the virus, such as changes in access to health care services, may only be reflected in patterns of mortality after several years.
  3. To best understand attribution of causes of death to excess mortality, detailed information on all causes is required. To date, all analysis of deaths from various diseases has included doctor certified deaths only as coronial investigations take time to finalise. Information on causes of death for coroner referred deaths registered in 2022 will not be published until later this year. 

While the second research question can offer important insights, there are still multiple issues to consider, including understanding how many deaths due to COVID-19 may be expected during a given year. There is not a clear answer to this. For example, when we consider how many COVID-19 associated deaths may be expected in Australia for the first three months of the year there have been marked differences each year. From January to March in 2021 there were four COVID-19 associated deaths. This compares to 3,658 for those months in 2022 and 1,639 in 2023. This results in a lack of stable time series for deaths due to the virus, introducing challenges for modelling expected counts. The ABS will work with stakeholders to develop a way forward, taking into account this issue. 

Considerations for interpreting information in this report

Outputs from excess mortality estimates will differ depending on the calculation applied and the scope of the input data. When interpreting the results in this report the following factors must be taken into consideration. 

  • This report highlights weeks where excess deaths are statistically significant. In any given period, even if no temporary health hazards influence the number of deaths (such as community transmission of influenza or COVID-19) there is some natural variation in patterns of mortality. While the actual number of deaths may be different from the expected number of deaths, it should fall within an expected range (i.e. there is a 95% chance that the expected number of deaths would lie between the upper and lower bounds of the confidence intervals). When actual observations (counts of death) exceed the upper threshold or drop below the lower threshold this indicates a statistically significant change in the pattern of mortality. This should be used in conjunction with the percentage excess mortality.
  • A single week above threshold does not necessarily suggest statistically significant excess mortality. Prolonged periods (2 or more weeks) where counts exceed thresholds suggest more strongly that the numbers of deaths are above or below normal. 
  • Excess mortality estimates for the Northern Territory are presented to the end of 2022 only. The Northern Territory has longer delays between date of death and date of registration. Reasons for this include a higher number of deaths referred to the coroner and geographical challenges including a high number of remote locations which can delay funeral times and access to services (and therefore delay the reporting of deaths).
  • Deaths occurring in the first quarter of 2023 are defined as those deaths occurring between 2 January 2023 and 26 March 2023. This is because this report produces data for deaths on a weekly basis using the ISO (International Organization for Standardisation) week date system. In this system, weeks are defined as seven-day periods which start on a Monday. Week 1 of any given year is the week which starts on the Monday closest to 1 January and for which the majority of its days fall in January.
  • Data is reported on the basis of occurrence, and is subject to revision as late registrations are received by the ABS. Only deaths that have been registered by 31 May 2023 and received by the ABS have been included in this publication. Please refer to the Timeliness and completeness of the data section of Provisional Mortality Statistics for further information. 
  • The section below presents a summary table with annual excess mortality numbers and percentages and graphs showing weekly mortality.
    • For Australia, two graphs are presented:

      • The first graph is a time series of all-cause deaths from January 2013 to March 2023 that has the number of deaths plotted against the expected number of deaths estimated from the regression. The upper and lower thresholds (1.96 standard errors, to give a 95% confidence interval) of the regression are also plotted. The 2013-2019 time series is shown to highlight the model fit used to estimate expected deaths in 2020-2023. 

      • The second graph focuses on 2020-2023, allowing closer inspection of patterns of death during the COVID-19 pandemic. In this graph, total deaths are presented with and without COVID-19 recorded on the death certificate. This provides an indication of how much COVID-19 contributed to excess mortality. Including this data series also highlights the importance of looking at all-cause mortality instead of only deaths due to the virus as there will be some deaths 'from' or 'with' COVID-19 where the virus is not recorded on the death certificate.

    • For jurisdictions, one graph is presented, focusing on 2020-2023. Key events are annotated, including the start of identified waves in Australia. While some jurisdictions did not experience these events (for example, Wave 2 was predominantly in Victoria), these annotations are still included so the impact of mortality during this time can be noted. 

Weekly all-cause mortality: Australia

  • Sustained statistically significant excess mortality continued to be recorded in Australia in the first quarter of 2023. 
  • For all weeks in the first quarter of 2023 the number of deaths exceeded the upper limit of usual variation reaching statistical significance (1,180 deaths). 
  • While excess mortality for 2023 for Australia is at 9.1% above expected, this is lower than excess mortality for the first quarter in 2022 which was 16.6% above expected. 
  • COVID-19 associated deaths were still a key contributor to excess mortality in Australia in January 2023. The contribution of COVID-19 to excess mortality decreased in February and March. 
  • Nearly every week in 2022 recorded higher than expected mortality. Mortality during this time exceeded the upper bounds, resulting in statistically significant excess mortality. 
  • COVID-19 was the main contributor to excess mortality during 2022. Excess mortality during this period corresponded with the Omicron waves. 
  • Deaths are significantly lower than expected from the week beginning 1 June to mid-July 2020, dropping below lower thresholds. Winter months are typically associated with higher mortality. These decreases provide insights into how public health measures put in place to manage the COVID-19 pandemic impacted mortality. 
  • Australia has recorded excess mortality in previous years, including 2014, 2015 and 2017. While the graph does not provide official excess mortality estimates for this time period, it shows an actual pattern of mortality where deaths were higher than expected. In past years excess mortality has typically occurred during the winter months and has been associated with virulent influenza seasons. 
Excess mortality by year, Australia, 2020-23
 ExpectedActualExcess% ExcessDeaths below usual variationDeaths above usual variationActual w/o reported COVID-19
2020170,006164,778-5,228-3.11,9559163,862
2021169,340171,7182,3781.40355170,274
2022171,692190,32618,63410.9010,587177,148
202336,10139,3813,2809.101,18037,861
  1. Data is provisional and subject to change.
  2. Years are based on a sum of ISO weeks derived from the weekly modelling. There are 53 weeks in 2020. There are 52 weeks in 2021 and 2022. Excess mortality has been estimated for the first 12 weeks of 2023. 
  3. Actual without reported COVID-19 has removed deaths 'from' or 'with' COVID-19 as recorded on the death certificate from all-cause mortality. 
  4. Deaths above and below the usual limits of variation are the number of deaths which exceed the upper and lower bounds. This signifies statistical significance. 
  5. Deaths in 2023 are deaths that occurred by 26 March and were registered and received by the ABS by 31 May 2023.
  1. Dates for key events are indicative only and may differ to other sources. 
  2. Data is provisional and will change as additional death registrations are received.
  3. The baseline period generating the regression model is from 2013 to 2019. The prediction window begins in 2020. 
  4. Data includes all deaths occurring by 26 March 2023 and registered and received by the ABS by 31 May 2023. 
  1. Dates for key events are indicative only and may differ to other sources. 
  2. Data is provisional and will change as additional death registrations are received.
  3. The baseline period generating the regression model is from 2013 to 2019. The prediction window begins in 2020. 
  4. Data includes all deaths occurring by 26 March 2023 and registered and received by the ABS by 31 May 2023. 

Weekly all-cause mortality: New South Wales

  • Excess mortality in New South Wales was 8.6% above expected for the first quarter of 2023. 
  • From mid-February 2023 deaths in New South Wales generally fall into the expected range of variation, i.e. there was no statistically significant excess mortality. 
  • Excess mortality for the first quarter of 2023 (8.6% above expected) was lower than the corresponding quarter in 2022 (20.5%).  
  • COVID-19 associated mortality has been a significant contributor to excess mortality in New South Wales during 2022 and early 2023. 
  • Sustained excess mortality started being recorded in New South Wales from the end of 2021 until the end of August 2022. Most deaths occurred during the Omicron period. 
  • Excess mortality was 10.7% above expected for 2022.
  • New South Wales did not record excess mortality during the Delta wave in 2021. 
  • Although most jurisdictions recorded lower than expected mortality across the winter and spring months of 2020. New South Wales was the only jurisdiction where this decrease was below the lower bounds for the entire winter period. 
Excess mortality by year, New South Wales, 2020-23
 ExpectedActualExcess% ExcessDeaths below usual variationDeaths above usual variationActual w/o reported COVID-19
202055,83753,552-2,285-4.16801353,488
202155,52655,582560.1308054,915
202256,08262,0745,99210.702,72757,280
202311,54012,5299898.6015212,014
  1. Data is provisional and subject to change. 
  2. Years are based on a sum of ISO weeks derived from the weekly modelling. There are 53 weeks in 2020. There are 52 weeks in 2021 and 2022. Excess mortality has been estimated for the first 12 weeks of 2023. 
  3. Actual without reported COVID-19 has removed deaths 'from' or 'with' COVID-19 as recorded on the death certificate from all-cause mortality. 
  4. Deaths above and below the usual limits of variation are the number of deaths which exceed the upper and lower bounds. This signifies statistical significance. 
  5. Deaths in 2023 are deaths that occurred by 26 March and were registered and received by the ABS by 31 May 2023.  
  1. Dates for key events are indicative only and may differ to other sources. 
  2. Data is provisional and will change as additional death registrations are received.
  3. The baseline period generating the regression model is from 2013 to 2019. The prediction window begins in 2020. 
  4. Data includes all deaths occurring by 26 March 2023 and registered and received by the ABS by 31 May 2023.

Weekly all-cause mortality: Victoria

  • Excess mortality in Victoria was 12.0% above expected for the first quarter of 2023. 
  • From mid-March, mortality falls into the expected range of variation. 
  • Excess mortality for the first quarter of 2023 (12.0% above expected) was lower than the corresponding quarter in 2022 (21.7%). 
  • Victoria had the second highest excess mortality recorded in 2022 at 13.2% above expected (second to Tasmania at 13.6%). Excess mortality in Victoria in 2022 was sustained, often above the upper bound. There were only 4 weeks in Victoria in 2022 where mortality was lower than expected. 
  • The highest peak in January (1,043 deaths) is similar to the highest peak in July (1,053 deaths). Typically winter has much higher rates of mortality than summer months. 
  • Victoria was the only jurisdiction to record statistically significant excess mortality during Wave 2 of the pandemic. Excess mortality during this period was driven by COVID-19 mortality. All other jurisdictions had low rates of infections. 
Excess mortality by year, Victoria, 2020-23
 ExpectedActualExcess% ExcessDeaths below usual variationDeaths above usual variationActual w/o reported COVID-19
202041,97541,608-367-0.916010640,794
202141,49642,9131,4173.4020742,158
202241,83247,3685,53613.202,34643,924
20238,7899,8401,0511202269,504
  1. Data is provisional and subject to change.
  2. Years are based on a sum of ISO weeks derived from the weekly modelling. There are 53 weeks in 2020. There are 52 weeks in 2021 and 2022. Excess mortality has been estimated for the first 12 weeks of 2023. 
  3. Actual without reported COVID-19 has removed deaths 'from' or 'with' COVID-19 as recorded on the death certificate from all-cause mortality. 
  4. Deaths above and below the usual limits of variation are the number of deaths which exceed the upper and lower bounds. This signifies statistical significance. 
  5. Deaths in 2023 are deaths that occurred by 26 March and were registered and received by the ABS by 31 May 2023.
  1. Dates for key events are indicative only and may differ to other sources. 
  2. Data is provisional and will change as additional death registrations are received.
  3. The baseline period generating the regression model is from 2013 to 2019. The prediction window begins in 2020. 
  4. Data includes all deaths occurring by 26 March 2023 and registered and received by the ABS by 31 May 2023.

Weekly all-cause mortality: Queensland

  • Excess mortality was 7.8% above expected in Queensland for the first quarter of 2023.
  • Although mortality has been above the expected number from October 2022, there is only a 3 week period from mid-December 2022 to early January 2023 where excess mortality is sustained above the upper bound (i.e. statistically significant). 
  • Excess mortality for the first quarter of 2023 (7.8% above expected) was lower than the corresponding quarter in 2022 (12.1%) 
  • Queensland had two distinct peaks of excess mortality in 2022 - one in the first 3 months of the year and a more severe wave during winter.
  • COVID-19 was a key contributor to excess mortality in 2022 and has remained so in the first quarter of 2023. 
  • Until the Omicron period, mortality in Queensland during 2021 fell within the expected range of variation. Excess mortality was less than 1% in 2021. 
  • Queensland was one of three jurisdictions (alongside New South Wales and the Australian Capital Territory) to record negative excess mortality in 2020 of more than 4%. 
Excess mortality by year, Queensland, 2020-23
 ExpectedActualExcess% ExcessDeaths below usual variationDeaths above usual variationActual w/o reported COVID-19
202033,79432,328-1,466-4.31813832,324
202134,02634,3102840.805934,306
202234,90038,4223,52210.101,24236,096
20237,5508,1415917.80577,820
  1. Data is provisional and subject to change. 
  2. Years are based on a sum of ISO weeks derived from the weekly modelling. There are 53 weeks in 2020. There are 52 weeks in 2021 and 2022. Excess mortality has been estimated for the first 12 weeks of 2023. 
  3. Actual without reported COVID-19 has removed deaths ‘from’ or ‘with’ COVID-19 as recorded on the death certificate from all-cause mortality. 
  4. Deaths above and below the usual limits of variation are the number of deaths which exceed the upper and lower bounds. This signifies statistical significance. 
  5. Deaths in 2023 are deaths that occurred by 26 March and were registered and received by the ABS by 31 May 2023.
  1. Dates for key events are indicative only and may differ to other sources. 
  2. Data is provisional and will change as additional death registrations are received.
  3. The baseline period generating the regression model is from 2013 to 2019. The prediction window begins in 2020. 
  4. Data includes all deaths occurring by 26 March 2023 and registered and received by the ABS by 31 May 2023.

Weekly all-cause mortality: South Australia

  • Excess mortality in South Australia was 8.9% above expected for the first quarter of 2023.
  • The number of deaths was within the expected range of variation for all weeks in the first quarter of 2023. 
  • Excess mortality for the first quarter of 2023 (8.9% above expected) was lower than the corresponding quarter in 2022 (12.2%)
  • Except for two weeks in March, excess mortality was sustained in South Australia from mid-January 2022 until the end of August 2022. COVID-19 was a key contributor to increased mortality during this time.  
  • Mortality was generally in the expected range of variation during the first two years of the pandemic. From June 2020 to the start of 2021, mortality was mostly lower than expected.  
Excess mortality by year, South Australia, 2020-23
 ExpectedActualExcess% ExcessDeaths below usual variationDeaths above usual variationActual w/o reported COVID-19
202014,41813,963-455-3.251913,959
202114,33914,406670.542114,402
202214,53215,8741,3429.2023914,824
20233,0383,3082708.9003,169
  1. Data is provisional and subject to change. 
  2. Years are based on a sum of ISO weeks derived from the weekly modelling. There are 53 weeks in 2020. There are 52 weeks in 2021 and 2022. Excess mortality has been estimated for the first 12 weeks of 2023. 
  3. Actual without reported COVID-19 has removed deaths ‘from’ or ‘with’ COVID-19 as recorded on the death certificate from all-cause mortality. . 
  4. Deaths above and below the usual limits of variation are the number of deaths which exceed the upper and lower bounds. This signifies statistical significance. 
  5. Deaths in 2023 are deaths that occurred by 26 March and were registered and received by the ABS by 31 May 2023.
  1. Dates for key events are indicative only and may differ to other sources. 
  2. Data is provisional and will change as additional death registrations are received.
  3. The baseline period generating the regression model is from 2013 to 2019. The prediction window begins in 2020. 
  4. Data includes all deaths occurring by 26 March 2023 and registered and received by the ABS by 31 May 2023. 

Weekly all-cause mortality: Western Australia

  • Excess mortality in Western Australia was 6.1% above expected in the first quarter of 2023.
  • While only two weeks exceeded the upper bound (reach statistical significance), mortality was consistently above the expected number of deaths in the first quarter of 2023. 
  • Excess mortality in Western Australia is similar in the first quarter of 2023 (6.1% above expected), compared to the first quarter of 2022 (6.3%). This is most likely related to the later opening of borders for travellers to WA which led to a later peak of COVID-19 infections. 
  • The first peak of excess mortality in WA occurred in the week ending 8 May 2022. This differed from most other jurisdictions where the first peak of excess mortality in 2022 occurred in January or February. 
  • From mid-March 2022, COVID-19 was a key contributor to excess mortality. This is notable across the winter months. 
  • During the first two years of the pandemic, mortality in WA was generally within the expected range of variation. There was less than 1% excess mortality in 2021. In 2020, lower than expected mortality was recorded, with a period of statistically significant lower than expected mortality between August and September. 
Excess mortality by year, Western Australia, 2020-23
 ExpectedActualExcess% ExcessDeaths below usual variationDeaths above usual variationActual w/o reported COVID-19
202015,89115,274-617-3.9261015,263
202115,96116,059980.601116,059
202216,33617,3451,0096.205916,409
20233,4733,6862136.10133,551
  1. Data is provisional and subject to change. 
  2. Years are based on a sum of ISO weeks derived from the weekly modelling. There are 53 weeks in 2020. There are 52 weeks in 2021 and 2022. Excess mortality has been estimated for the first 12 weeks of 2023. 
  3. Actual without reported COVID-19 has removed deaths ‘from’ or ‘with’ COVID-19 as recorded on the death certificate from all-cause mortality. 
  4. Deaths above and below the usual limits of variation are the number of deaths which exceed the upper and lower bounds. This signifies statistical significance. 
  5. Deaths in 2023 are deaths that occurred by 26 March and were registered and received by the ABS by 31 May 2023. 
  1. Dates for key events are indicative only and may differ to other sources. 
  2. Data is provisional and will change as additional death registrations are received.
  3. The baseline period generating the regression model is from 2013 to 2019. The prediction window begins in 2020. 
  4. Data includes all deaths occurring by 26 March 2023 and registered and received by the ABS by 31 May 2023. 

Weekly all-cause mortality: Tasmania

  • Excess mortality in Tasmania was 17.3% above expected for the first quarter of 2023. Tasmania has the highest excess mortality of all jurisdictions. 
  • In the last two weeks of March, the number of deaths exceeded the upper range of the usual limits of variation. 
  • Tasmania is the only jurisdiction where excess mortality was higher in the first quarter of 2023 (17.3% above expected) than the first quarter of 2022 (12.3%). 
  • Excess mortality was sustained over the winter months in 2022 with COVID-19 contributing to higher mortality. 
  • Prior to September 2021, mortality in Tasmania was generally in the expected range. Similar to other states, there was a period of lower than expected mortality in 2020. 
Excess mortality by year, Tasmania, 2020-23
 ExpectedActualExcess% ExcessDeaths below usual variationDeaths above usual variationActual w/o reported COVID-19
20204,6314,462-169-3.61904,445
20214,5364,8012655.83144,801
20224,5105,12261213.60714,786
20239311,09216117.30191,039
  1. Data is provisional and subject to change. 
  2. Years are based on a sum of ISO weeks derived from the weekly modelling. There are 53 weeks in 2020. There are 52 weeks in 2021 and 2022. Excess mortality has been estimated for the first 12 weeks of 2023. 
  3. Actual without reported COVID-19 has removed deaths ‘from’ or ‘with’ COVID-19 as recorded on the death certificate from all-cause mortality. 
  4. Deaths above and below the usual limits of variation are the number of deaths which exceed the upper and lower bounds. This signifies statistical significance. 
  5. Deaths in 2023 are deaths that occurred by 26 March and were registered and received by the ABS by 31 May 2023.
  1. Dates for key events are indicative only and may differ to other sources. 
  2. Data is provisional and will change as additional death registrations are received.
  3. The baseline period generating the regression model is from 2013 to 2019. The prediction window begins in 2020. 
  4. Data includes all deaths occurring by 26 March 2023 and registered and received by the ABS by 31 May 2023.

Weekly all-cause mortality: Northern Territory

  • Excess mortality estimates for the Northern Territory are only available until the end of 2022. There is a longer time between the date of death and date of registration in the NT, meaning there is a delay in reporting more complete figures. This is due to the dispersed geography, with the wet season and remote communities sometimes leading to a delay in burial and funerals (and therefore death registrations). Additionally, a higher proportion of deaths are referred to a coroner in the NT. 
  • While the NT recorded 10.1% above expected excess mortality in 2022, this number should be treated with caution. There are a small number of deaths occurring in the NT each year, and often there are less than 20 deaths per week. The small number of deaths in the NT is due to a combination of factors, including the Territory's smaller population size and younger population (Darwin has the youngest median age of all capital cities). 
  • Due to the low number of weekly deaths, calibration by age was not possible. Predicting an expected number of deaths from a small base can lead to volatility. This is evident from the range of the expected number - the upper and lower bounds are wider to account for this high variation. 
  • There are some weeks in 2022 where mortality was above what was expected for sustained periods, indicating some likely excess mortality. For most of the pandemic, mortality in the NT has been in the expected range of variation. 
  • COVID-19 contributed to increased mortality in 2022. While numbers are small, COVID-19 mortality peaks occurred at similar times to other jurisdictions, with an increase in February and March and then again over the winter months. 
Excess mortality by year, Northern Territory, 2020-22
 ExpectedActualExcess% ExcessDeaths below usual variationDeaths above usual variationActual w/o reported COVID-19
20201,1431,160171.5021,160
20211,1201,196766.8071,195
20221,1221,24111910.6121,156
  1. Data is provisional and subject to change. 
  2. Years are based on a sum of ISO weeks derived from the weekly modelling. There are 53 weeks in 2020. There are 52 weeks in 2021 and 2022. 
  3. Actual without reported COVID-19 has removed deaths 'from' or 'with' COVID-19 as recorded on the death certificate from all-cause mortality. 
  4. Deaths above and below the usual limits of variation are the number of deaths which exceed the upper and lower bounds. This signifies statistical significance. 
  1. Dates for key events are indicative only and may differ to other sources. 
  2. Data is provisional and will change as additional death registrations are received.
  3. The baseline period generating the regression model is from 2013 to 2019. The prediction window begins in 2020. 
  4. Data includes all deaths occurring by December 2022 and registered and received by the ABS by 31 May 2023. 

Weekly all-cause mortality: Australian Capital Territory

  • Excess mortality in the Australian Capital Territory was 8.9% above expected for the first quarter of 2023.
  • Mortality in all weeks except one in the first quarter of 2023 was within the expected bounds of variation. 
  • Excess mortality for the first quarter of 2023 (8.9% above expected) was similar to the first quarter of 2022 (9.2%).   
  • Over the winter months of 2022, COVID-19 was a key contributor to excess mortality.  
  • Mortality was lower than expected across the first two years of the pandemic in the Australian Capital Territory. 
  • Similar to the Northern Territory, the population of the ACT is small, with Canberra having the second youngest population in Australia. This results in a small number of deaths by week and the predicted count can be subject to volatility. 
Excess mortality by year, Australian Capital Territory, 2020-23
 ExpectedActualExcess% ExcessDeaths below usual variationDeaths above usual variationActual w/o reported COVID-19
20202,5392,431-108-4.3202,429
20212,5212,451-70-2.8002,438
20222,5692,88031112.10112,673
2023550599498.904582
  1. Data is provisional and subject to change. 
  2. Years are based on a sum of ISO weeks derived from the weekly modelling. There are 53 weeks in 2020. There are 52 weeks in 2021 and 2022. Excess mortality has been estimated for the first 12 weeks of 2023. 
  3. Actual without reported COVID-19 has removed deaths 'from' or 'with' COVID-19 as recorded on the death certificate from all-cause mortality. 
  4. Deaths above and below the usual limits of variation are the number of deaths which exceed the upper and lower bounds. This signifies statistical significance. 
  5. Deaths in 2023 are deaths that occurred by 26 March and were registered and received by the ABS by 31 May 2023.  
  1. Dates for key events are indicative only and may differ to other sources. 
  2. Data is provisional and will change as additional death registrations are received.
  3. The baseline period generating the regression model is from 2013 to 2019. The prediction window begins in 2020.  
  4. Data includes all deaths occurring by 26 March 2023 and registered and received by the ABS until 31 May 2023. 

Methodology

The analysis of 2020-2023 mortality data is based on a model developed by Serfling¹⁰ and later adapted by the US Center for Disease Control (CDC) and New South Wales Health (NSW Health). This section provides an overview of how the model has been developed over time, key aspects of the model, and how the model has been adapted and applied by the ABS in this analysis.

Historical development of model

Historically this type of model has been used to estimate excess mortality caused by influenza.

As early as 1932, Collins determined that excess mortality during winter months in the United States of America (USA) was a consequence of epidemic influenza and therefore could be used as an indicator for the recognition of influenza outbreaks.²

In 1963, Serfling described a cyclic regression model, based on the seasonal pattern of pneumonia and influenza (P&I) deaths, to infer excess deaths due to influenza.¹⁰

Since then, models based on Serfling's have been applied in a number of temperate countries (USA,³,⁴ France,⁴ Australia,⁴ and Italy⁵) to demonstrate that excess mortality occurring during winter months is associated with the pandemic and seasonal epidemics of influenza.

The CDC took the approach further and established the 122 Cities Surveillance System to provide timely, prospective information on excess mortality due to influenza. ⁶,⁷

The New South Wales mortality surveillance was included as part of the routine Influenza Surveillance Program prior to COVID-19 and used an adaptation of the Serfling model to monitor for influenza epidemics. This system primarily monitored the proportion of influenza positive specimens among all respiratory specimens from the major public health laboratories, the proportion of emergency department visits diagnosed with influenza, and outbreaks of influenza reported to Public Health Units by residential care facilities.⁹ 

Aspects of the Serfling model and its implementations

The Serfling model uses historical data to predict current patterns of mortality. This enables the identification of any deviation from that prediction that may signal an influenza epidemic. In developing this model, Serfling recognised that past influenza epidemics could overly influence the regression, limiting the ability of the model to identify any new epidemic. For this reason, Serfling excluded past epidemics when fitting the model. 

An important feature of this methodology is the identification of a threshold that can be used as the upper and/or lower limit for the data. Observations that lie between these upper and lower limits are considered to be normal. The chosen threshold will vary depending on the desired specificity of the analysis and the methods used for calculation. In statistics there are default or familiar thresholds used in measuring these boundaries such as 1.96 standard errors either side of the mid-point.

An upper threshold was chosen to define the upper limit of the expected number of deaths in the absence of an influenza epidemic. This upper threshold was calculated as the predicted number of deaths in a given week plus a constant multiple of the standard error of the time series (the differences between each value predicted by the model and the actual observed values, or the 'model residuals'). Serfling chose the constant multiple to be 1.64 standard errors, and considered 2 consecutive weeks above the threshold to indicate epidemic behaviour.¹⁰

The NSW Health model used for the Influenza Surveillance Program is based on the Serfling model. This variant of the model fits a cyclic linear regression to data for the previous 5 calendar years and then forecasts the current year's time series from data up to the end of the previous year.

Data source

Refer to the methodology section of Provisional Mortality Statistics for a description of the scope of data used in this model.

ABS adaptation of the model

The ABS has applied key characteristics of the model used in the New South Wales Influenza Surveillance Program. The ABS model is also based on the application of a cyclic linear regression model to the time series of weekly age-specific death rates. The model has been applied to both doctor certified and coroner referred deaths from all causes. 

The approach used by the ABS involves use of a 'robust' estimation procedure for fitting the model. Epidemics can cause increased mortality which leads to outliers in time series data. By down-weighting outliers it aims to fit the model to patterns of mortality that are expected to occur in the absence of an epidemic. This robust regression down-weights the influence of extreme observations (outliers) and is applied to a seven-year baseline time series (weeks commencing on a Monday). Some additional adjustments were made to extreme outliers, for example the 2017 influenza season. This adjustment involved applying a missing value to some age groups across some weeks where the actual rate was unusually high. This was done separately for each jurisdiction. 

The cyclical regression model includes: a linear time term, t, with values 1, 2, 3, ... for each week of the time series. Also included are annual seasonal harmonic variables to describe the cyclical seasonal background pattern. The harmonic variables are functions of the week number, t, and the periodicity in the same units – in this case, yearly (52 weeks). The 2 harmonic variables in this case are: sine(2π t/52) and cosine(2π t/52). For years with 53 weeks these harmonic terms have been divided by 53 instead of 52.

The final model was:

\(Expected(proportion)=A+Bt+Csin(2πt/52)+Dcos(2πt/52)\)

where A, B, C and D, are the coefficients calculated from the regression. Previous versions of this model also include a time squared term. This term was removed as it resulted in increasingly unrealistic output as the prediction window extended further from the modelling window.

To evaluate the approach, the model was fitted using PROC ROBUSTREG in SAS Software with the simplest, default 'M estimation' method.¹² The PROC SCORE procedure was then used to forecast values for 2020-2023. The standard error and threshold is derived from the stdi option in PROC ROBUSTREG which is run a second time with the 2020-2023 results attached.

Threshold identification

Identifying the threshold at which to signal excess mortality will vary depending on the goals of the research, application of the method, and seasonality of diseases of interest. Several national statistical organisations are using different models including Farrington surveillance algorithms, z -scores and prediction intervals to inform such thresholds.

A method for calculating the threshold for this analysis was selected which had a broader application and was more easily interpreted across a range of conditions, namely 1.96 standard errors or a 95% confidence interval. Use of the 95% confidence interval does introduce the expectation that random weekly variation could lead to counts that exceed thresholds 5% of the time. As such, individual weeks that exceed thresholds should be interpreted with caution. However, any prolonged period of weeks, 2 or more, exceeding the expected range can suggest excess mortality has been signalled.

The SAS function PROC ROBUSTREG offers the STDI parameter to calculate the 'standard error of prediction' of the model¹² which is used to define 1 standard error. The standard error of prediction is a more logical choice for assigning the threshold of excess mortality than the root mean square error (standard error of the model residuals) because it incorporates not only the variance of the residuals but also the variance of the model parameter estimates. This provides an estimate of the expected variability of the observed values in the absence of an epidemic.

Determining the baseline

Choosing the reference period (i.e. the number of years in the baseline) is important as it can change the expected number of deaths. When selecting the reference period key attributes were required:

  • There needed to be enough input available to predict the number of deaths from 2020-2023. As the paper is looking at excess mortality in the absence of the pandemic, no data from 2020-2023 is part of the reference period when predicting the expected number of deaths. Estimates of excess mortality are more accurate for years that are closer to the reference period.
  • A stable and clear mortality trend needed to be identified. This was particularly important as the model used is a harmonic with trend, meaning the pattern of death should be stable over a period to estimate an expected trend accurately.
  • The baseline period needed to be applied consistently across jurisdictions.

To make the decision, a sensitivity analysis was conducted. This analysis tested three reference periods: 2010-2019, 2013-2019 and 2015-2019. Standardised death rates were used to determine which years to test (these are presented below). Each baseline produced slightly different expected counts of deaths, which altered the % excess figures. The ABS decided to implement 2013-2019 as the predictor reference period. This baseline was chosen because:

  • There was a large decline in mortality between 2017 and 2018. This is likely due to the severe influenza season in 2017 causing some mortality displacement in 2018. Mortality displacement is an epidemiological concept which describes the phenomenon of a period of very high mortality being followed by a period of low mortality. Even controlling for 2017, the model was overcompensating for the rate of decline during 2015-2019, resulting in a very low number of expected deaths in 2022 and 2023.
  • Not all jurisdictions experienced a severe influenza season in 2017. Western Australia for example, had higher mortality rates in 2015 and 2016. Variability across jurisdictions meant that 2015-2019 had different outcomes across jurisdictions.
  • Western Australia had steeper declines in mortality rates between 2015-2016 and 2017 compared to other jurisdictions. Starting the baseline at the highest mortality points was overstating the rate of decline in WA and resulting in a very low number of expected deaths in 2022 and 2023. This was also affected by the model – as a harmonic with trend model is being used, volatility in the input trend affected the expected number of deaths.
  • For smaller jurisdictions with low numbers of weekly deaths, adding additional years onto the baseline provided a more stable trend.
  • There was some excess mortality in 2014, 2015 and 2017. Adjustments were made to outliers to control for these. 2013 was a year of stable mortality where no adjustments had to be made across any jurisdictions.
  • 2010-2019 and 2013-2019 produced similar results. However, there is less population change to account for from 2013 onwards. This was especially important for smaller jurisdictions where age adjustment was not as precise due to small numbers.
  • This model will also be used for analysis of diseases. There were a number of coding changes in 2013. Starting the reference period ensures continuity of time series for this analysis.

Modelling rates

Weekly mortality rates were calculated for 2013-2019 to predict expected weekly rates of death for 2020-2023. Rates were then converted to numbers for ease of interpretation.

Age groups selected for modelling ASDRs went through a number of assessments including:

  • A consistent number of deaths each week, ideally at least 20 deaths per week in the age group.
  • Population to be used as denominators should have at least 30 people in each age group.

Final age groups vary by jurisdiction based on the above criteria. Where possible, age groups were modelled separately up to 95 years and over (Australia, New South Wales, Victoria and Queensland). For South Australia and Western Australia age groups were modelled up to 90 years and over. In Tasmania age groups were modelled to 85 years and over. Weekly numbers of deaths for the Northern Territory and the Australian Capital Territory were too small to allow robust age-specific rates to be calculated, and crude rates (total population) were used instead.

Consideration was given to whether age-specific rates or age-standardised rates should be modelled. Both age-specific rates and age-standardised rates control for the age composition of a population and take into account growth rates within a population over time. Age-specific rates show the intensity of deaths within a population and express real-life mortality and population loss. Age-standardised rates are a modelled rate, standardised against a hypothetical population to enable comparison across cohorts. As comparison was not the main focus, age-specific rates were chosen as the model input.

References

1. CDC. Excess Deaths Associated with COVID-19 https://www.cdc.gov/nchs/nvss/vsrr/covid19/excess_deaths.htm#techNotes

2. Collins SD. Excess mortality from causes other than influenza and pneumonia during influenza epidemics. Public Health Rep 1932;47:2159–2180.

3. Reichert T, Simonsen L, Sharma A, Pardo S, Fedson D, Miller M. Influenza and the winter increase in mortality in the United States, 1959–1999. Am J Epidemiol 2004;160:492–502.

4. Viboud C, Boelle P, Pakdaman K, Carrat F, Valleron A, Flahault A. Influenza epidemics in the United States, France, and Australia, 1972–1997. Emerg Infect Dis 2004;10:32–39.

5. Rizzo C, Bella A, Viboud C, Simonsen L, Miller M, Rota M, et al. Trends for influenza-related deaths during pandemic and epidemic seasons, Italy, 1969–2001. Emerg Infect Dis 2007;5:694–699.

6. Centers for Disease Control and Prevention. 122 Cities Mortality Reporting System, Manual of Procedures. Atlanta (Georgia, United States): US Department of Health and Human Services; 2004.

7. Simonsen L, Clarke MJ, Stroup DF, Williamson GD, Arden NH, Cox NJ. A method for timely assessment of influenza-associated mortality in the United States. Epidemiology 1997;8:90–395.

8. O'Brien K, Barr IG. Annual Report of the National Influenza Surveillance Scheme, 2006. Commun Dis Intell 2007;31:167–179.

9. NSW Health, Communicable Diseases Branch. New South Wales Influenza Surveillance Report. NSW Health.

10. Serfling RE. Methods for current statistical analysis of excess pneumonia-influenza deaths. Public Health Rep 1963;78:494–506.

11. The ROBUSTREG procedure. SAS/STAT(R) 12.3 User's Guide. Cary (USA): SAS Institute, 2013. Available from: https://support.sas.com/documentation/onlinedoc/stat/123/rreg.pdf

12. Model Fit and Diagnostic Statistics. SAS/STAT(R) 9.2 User's Guide. Cary (USA): SAS Institute, 2008.

Data downloads

Excess mortality, Australia and by state, Jan 2013 - Mar 2023

Data files
Download xlsx [471.34 KB]