Trends and geographic differences in life expectancy in Australia during the COVID-19 pandemic: a multiple cause of death analysis Authors: Tim Adair, Sergey Timonin, Vladimir Canudas-Romo Published: 2026 ABSTRACT BackgroundDuring the pandemic, Australia experienced relatively stringent restrictions and a distinctive life expectancy trend.This study assesses the contribution of multiple causes of death to these trends, as well as geographic differences in life expectancy, during the pandemic across Australia.MethodsData on deaths by age, sex, cause and capital city classification in Australian death registration data from 2017 to 2023 were accessed from the Australian Bureau of Statistics. We estimated life expectancy at birth and applied decomposition techniques to quantify the contribution of multiple causes of death to changes in life expectancy, including by capital city. Multiple cause of death analysis measured whether a death involved respiratory infections and/or COVID-19, and the accompanying underlying cause of death.ResultsOver half the 0.6 years life expectancy increase from 2017-19 to 2020 was from fewer deaths involving respiratory infections such as pneumonia, especially at age 80 years and above. Following a small life expectancy decline from 2020 to 2021, most of the 0.7 years decline in life expectancy in from 2021 to 2022 resulted from large increases in COVID-19 mortality following relaxation of restrictions. Deaths with a respiratory infection but not involving COVID-19 mostly had a non-communicable disease such as cardiovascular disease as the underlying cause. All geographic regions mirrored national trends by experiencing life expectancy increases in 2020 (especially from lower respiratory infection mortality) and declines in 2022 due to rising COVID-19 deaths, with the worst trends in Melbourne.ConclusionsAustralia experienced large fluctuations in life expectancy during the COVID-19 pandemic that were unique compared with many other high-income countries. The life expectancy increase in 2020 was mostly due to reduced circulation of respiratory infections like pneumonia during restrictions. However, upon cessation of restrictions in 2022, a large increase in COVID-19 mortality led to substantial life expectancy falls. Life expectancy trends did not vary greatly by capital city because restrictions were effective at preventing further deaths from COVID-19 after local outbreaks. Multiple cause of death data are very useful to understand the interaction of COVID-19, respiratory infections and non-communicable diseases during pandemic and the role of pandemic-related restrictions.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12963-026-00490-5. Introduction Australia’s life expectancy exhibited a distinctive trend during the COVID-19 pandemic. It was one of the few high-income countries where life expectancy increased sharply in 2020, when relatively stringent restrictions were introduced by Australia’s Federal, State, and Territory Governments to control the spread of COVID-19 [1–5]. These restrictions were applied to gatherings and visits to hospitals and aged care centres and involved closures of international and state/territory borders, schools and non-essential businesses [6]. The restrictions varied geographically, with the second-largest city of Melbourne having the most stringent restrictions during 2020 and, along with the largest city Sydney and the capital Canberra, in 2021, in response to the rapid spread of COVID-19 in these locations [6]. National life expectancy declined sharply in 2022 as there was a large rise in COVID-19 deaths (with the Omicron wave) once restrictions were relaxed throughout Australia after vaccination targets were reached [3]. A timeline of Australia’s COVID-19 restrictions is presented in Additional File 1. The introduction of pandemic-related restrictions not only helped to prevent the spread of COVID-19 but also reduced the circulation of (mostly) airborne respiratory infections such as pneumonia and influenza [7, 8]. It can therefore be hypothesised that lower mortality involving respiratory infections due to pandemic restrictions would have contributed to a significant proportion of the increase in life expectancy in 2020, while higher mortality involving COVID-19 would have contributed to much of the fall in life expectancy in 2022 following relaxation of restrictions. Trends in remaining deaths that involved neither respiratory infections nor COVID-19 may have been related to the indirect impacts of lockdowns and their relaxation, such as mortality from injuries (e.g. road traffic accidents), and factors unrelated to the pandemic. To understand the contribution of cause-specific mortality to such changes in life expectancy, both academic research and official statistics almost always focus on measuring the underlying cause of death [3, 9]. The underlying cause is a single cause that initiated the chain of events leading directly to death. It is determined by the causes listed on the death certificate (International Form of Medical Certificate of Cause of Death) and the application of World Health Organization (WHO) International Classification of Diseases, 10th Revision (ICD-10) coding rules, which in Australia and some other high-income countries is implemented using the automated coding software Iris [3, 4, 10–12]. In Australia, an underlying cause analysis showed that the increase in life expectancy in 2020 was due to declines in mortality from several non-communicable diseases, with only a slight negative contribution from COVID-19 mortality [3]. However, measurement of the contribution of just the underlying cause to life expectancy trends limits understanding of the role of cause-specific mortality during the pandemic. Pneumonia, for instance, is commonly reported on the death certificate as an immediate cause of death with another condition, mostly a non-communicable disease, being the underlying cause. Hence, in an underlying cause analysis, an increase in life expectancy caused by reduced infection from and mortality involving community-acquired pneumonia during pandemic restrictions would be shown as due largely to lower non-communicable disease mortality. COVID-19, however, is commonly an underlying cause of death because it is mostly reported in Part 1 of the death certificate, i.e., in the chain of events leading directly to death; when COVID-19 is reported in Part 1, ICD-10 coding rules stipulate it to be the underlying cause [13]. In contrast, COVID-19 is less commonly reported in Part 2 as another significant condition contributing to the death but not causing it, in which case it is not the underlying cause of death. More information about COVID-19 as a cause of death is shown in the Additional File 1. This paper uses multiple cause of death (MCOD) data in Australia to measure the contribution of causes of death to changes in life expectancy. It focuses on COVID-19 and respiratory infections such as pneumonia and influenza, based on whether they were reported anywhere on the death certificate (i.e. a MCOD). We compare the analysis of COVID-19 and respiratory infections (MCOD) to solely the underlying cause and reconcile differences between the two approaches by measuring the underlying causes for deaths involving each of COVID-19, respiratory infections and other causes. Life expectancy trends and the role of MCODs are analysed by capital city classification (Greater Sydney, Greater Melbourne, other capital cities, outside capital cities) to assess whether there was a differential impact given that pandemic restrictions varied by location. These analyses will help understand the interaction of COVID-19 and respiratory infections with non-communicable diseases and more generally the unique mortality dynamics that prevailed in Australia during the pandemic. Methods We used Australian death registration data from 2017 to 23 available in the Australian Bureau of Statistics (ABS) Person Level Integrated Data Asset (PLIDA) [14, 15]. We analysed all diseases and conditions reported on the death certificate, i.e. MCODs. Population data were available from the ABS [16]. Life expectancy at birth was estimated for each sex for 2017-19 (to reduce uncertainty in the reference pre-pandemic period) and each year from 2020 to 2023. Standard life expectancy decomposition techniques were used to measure the contribution of age and cause of death to changes in life expectancy at birth (see Additional File 1) [17, 18]. We calculated 95% confidence intervals using bootstrapping assuming a Poisson distribution of deaths. We aggregated the years 2020-23 to assess the overall change in life expectancy during the pandemic from 2017 to 19, including by capital city classification, given that there were sharp annual increases and decreases which may mask the overall change during the pandemic. However, we did not analyse causes of death for 2020-23 given our primary interest for causes are the annual fluctuations. Two analyses of causes of death were performed to measure their contribution to life expectancy change: We firstly used a MCOD approach (reported anywhere on the death certificate), comprising mutually exclusive categories of causes (see Fig. 1):Respiratory infections – pneumonia (ICD-10 codes J12-J18; over 90% of pneumonia reported on death certificates in our study is J18 “unspecified pneumonia”), influenza (J09-J11), other respiratory infections (J20-J22; acute bronchitis/ bronchiolitis, unspecified acute lower respiratory infection) – without COVID-19 (U07.1-U07.2). We did not consider Respiratory failure not elsewhere classified and Other respiratory disorders (J96-J99)Respiratory infections with COVID-19 COVID-19 without respiratory infections Neither respiratory infections nor COVID-19 (i.e. all other deaths)Secondly, we analysed the underlying causes for deaths involving the MCODs of: i) respiratory infections, ii) COVID-19 and iii) neither respiratory infections nor COVID-19 (we could not analyse four categories of MCOD as in 1a-d above because of an insufficient cases in some categories to meet the ABS output rules; hence, respiratory infections and COVID-19 are not mutually exclusive in this classification). That is, the contribution to life expectancy change of each of these three MCODs (i-iii) was divided into nine underlying causes: We firstly used a MCOD approach (reported anywhere on the death certificate), comprising mutually exclusive categories of causes (see Fig. 1):Respiratory infections – pneumonia (ICD-10 codes J12-J18; over 90% of pneumonia reported on death certificates in our study is J18 “unspecified pneumonia”), influenza (J09-J11), other respiratory infections (J20-J22; acute bronchitis/ bronchiolitis, unspecified acute lower respiratory infection) – without COVID-19 (U07.1-U07.2). We did not consider Respiratory failure not elsewhere classified and Other respiratory disorders (J96-J99)Respiratory infections with COVID-19 COVID-19 without respiratory infections Neither respiratory infections nor COVID-19 (i.e. all other deaths) We firstly used a MCOD approach (reported anywhere on the death certificate), comprising mutually exclusive categories of causes (see Fig. 1):Respiratory infections – pneumonia (ICD-10 codes J12-J18; over 90% of pneumonia reported on death certificates in our study is J18 “unspecified pneumonia”), influenza (J09-J11), other respiratory infections (J20-J22; acute bronchitis/ bronchiolitis, unspecified acute lower respiratory infection) – without COVID-19 (U07.1-U07.2). We did not consider Respiratory failure not elsewhere classified and Other respiratory disorders (J96-J99)Respiratory infections with COVID-19 COVID-19 without respiratory infections Neither respiratory infections nor COVID-19 (i.e. all other deaths) Respiratory infections – pneumonia (ICD-10 codes J12-J18; over 90% of pneumonia reported on death certificates in our study is J18 “unspecified pneumonia”), influenza (J09-J11), other respiratory infections (J20-J22; acute bronchitis/ bronchiolitis, unspecified acute lower respiratory infection) – without COVID-19 (U07.1-U07.2). We did not consider Respiratory failure not elsewhere classified and Other respiratory disorders (J96-J99) Respiratory infections – pneumonia (ICD-10 codes J12-J18; over 90% of pneumonia reported on death certificates in our study is J18 “unspecified pneumonia”), influenza (J09-J11), other respiratory infections (J20-J22; acute bronchitis/ bronchiolitis, unspecified acute lower respiratory infection) – without COVID-19 (U07.1-U07.2). We did not consider Respiratory failure not elsewhere classified and Other respiratory disorders (J96-J99) Respiratory infections with COVID-19 Respiratory infections with COVID-19 COVID-19 without respiratory infections COVID-19 without respiratory infections Neither respiratory infections nor COVID-19 (i.e. all other deaths) Neither respiratory infections nor COVID-19 (i.e. all other deaths) Secondly, we analysed the underlying causes for deaths involving the MCODs of: i) respiratory infections, ii) COVID-19 and iii) neither respiratory infections nor COVID-19 (we could not analyse four categories of MCOD as in 1a-d above because of an insufficient cases in some categories to meet the ABS output rules; hence, respiratory infections and COVID-19 are not mutually exclusive in this classification). That is, the contribution to life expectancy change of each of these three MCODs (i-iii) was divided into nine underlying causes: Secondly, we analysed the underlying causes for deaths involving the MCODs of: i) respiratory infections, ii) COVID-19 and iii) neither respiratory infections nor COVID-19 (we could not analyse four categories of MCOD as in 1a-d above because of an insufficient cases in some categories to meet the ABS output rules; hence, respiratory infections and COVID-19 are not mutually exclusive in this classification). That is, the contribution to life expectancy change of each of these three MCODs (i-iii) was divided into nine underlying causes: Fig. 1Classification of causes of death (MCOD) Classification of causes of death (MCOD) Cardiovascular diseases (I00-I99)Dementia (F00-F03, G30-G31)Cancers (C00-D49)Chronic respiratory diseases (J40-J44)COVID-19 (U07.1-U07.2)Respiratory infections (J09-J18, J20-J22)Other non-communicable diseases (D50-E99, F04-F09, F17, F20-G29, G32-G99, J00-J08, J19, J23-J99, K00-K69, K71-K73, K75, K76.3, K77-N99, Q00-Q99), substance-related disorders (F10-F16, F18-F19, K70, K74, K76.0-K76.2, K76.4-K76.9, X40-X45, Y10-15) and other infections, maternal and perinatal conditions (A00-B99, O00-P99)Injuries (V00-X39, X46-Y09, Y16-Y99)Ill-defined causes (R00-R99) Cardiovascular diseases (I00-I99) Cardiovascular diseases (I00-I99) Dementia (F00-F03, G30-G31) Dementia (F00-F03, G30-G31) Cancers (C00-D49) Cancers (C00-D49) Chronic respiratory diseases (J40-J44) Chronic respiratory diseases (J40-J44) COVID-19 (U07.1-U07.2) COVID-19 (U07.1-U07.2) Respiratory infections (J09-J18, J20-J22) Respiratory infections (J09-J18, J20-J22) Other non-communicable diseases (D50-E99, F04-F09, F17, F20-G29, G32-G99, J00-J08, J19, J23-J99, K00-K69, K71-K73, K75, K76.3, K77-N99, Q00-Q99), substance-related disorders (F10-F16, F18-F19, K70, K74, K76.0-K76.2, K76.4-K76.9, X40-X45, Y10-15) and other infections, maternal and perinatal conditions (A00-B99, O00-P99) Other non-communicable diseases (D50-E99, F04-F09, F17, F20-G29, G32-G99, J00-J08, J19, J23-J99, K00-K69, K71-K73, K75, K76.3, K77-N99, Q00-Q99), substance-related disorders (F10-F16, F18-F19, K70, K74, K76.0-K76.2, K76.4-K76.9, X40-X45, Y10-15) and other infections, maternal and perinatal conditions (A00-B99, O00-P99) Injuries (V00-X39, X46-Y09, Y16-Y99) Injuries (V00-X39, X46-Y09, Y16-Y99) Ill-defined causes (R00-R99) Ill-defined causes (R00-R99) To assess how trends in life expectancy differed between geographic regions that experienced different pandemic restrictions, we analysed trends in life expectancy and the contribution of MCODs (analysis [1] above) by capital city classification (derived from the ABS Greater Capital City Statistical Areas): more populous cities of population above 5 million of (1) Greater Sydney and (2) Greater Melbourne, (3) other capital cities with smaller populations (Brisbane, Adelaide, Perth, Australian Capital Territory (Canberra)), and (4) outside capital cities (including the state of Tasmania and the Northern Territory) [19, ]. Results Life expectancy trends Australian life expectancy at birth rose by 0.57 years for males (95% confidence interval [CI] 0.48 to 0.66) and 0.57 years for females (0.50 to 0.66) from 2017 to 19 to 2020, fell moderately in 2021 and then declined sharply in 2022 by 0.70 years for males (-0.81 to -0.59) and 0.64 years for females (-0.74 to -0.54) to reach a level lower than in 2017-19 (Fig. 2; Additional File 2, Supplementary Table 1). In 2023 life expectancy rebounded by 0.75 years for males (0.64 to 0.86) and 0.62 years for females (0.52 to 0.72). Over 2020-23 combined, life expectancy was 0.27 (0.20 to 0.33) years higher than in 2017-19 for males and 0.18 years (0.13 to 0.23) higher for females. The largest contribution to annual fluctuations in life expectancy was from age 80 and above, especially from 2021 to 2022 and particularly for females, while from 2022 to 2023 there was a larger contribution in ages 0–79 years for males than females (Additional File 2, Supplementary Fig. 1). Fig. 2Life expectancy at birth, Australia, 2017-19 to 2023. Source: ABS [14–16] Life expectancy at birth, Australia, 2017-19 to 2023. Source: ABS [14–16] Contribution of MCODs Of the increase in life expectancy in 2020 compared with 2017-19, over half was due to lower mortality involving respiratory infections (i.e. as a MCOD) while the remainder contribution was from deaths with neither respiratory infections nor COVID-19 (Fig. 3). In 2022, the sharp decline in life expectancy was predominantly from higher mortality with COVID-19, with a significant proportion having a respiratory infection. Approximately half of the rise in 2023 was from lower mortality without COVID-19 nor respiratory infection, with almost all the remainder due to lower COVID-19 mortality. Fig. 3Contribution of MCODs and age groups to change in life expectancy, Australia, 2017-19 to 2023. Resp. inf.: Respiratory infections. LE: Life expectancy. A more detailed analysis by age group is shown in Additional File 2, Supplementary Fig. 2. Source: ABS [14–16] Contribution of MCODs and age groups to change in life expectancy, Australia, 2017-19 to 2023. Resp. inf.: Respiratory infections. LE: Life expectancy. A more detailed analysis by age group is shown in Additional File 2, Supplementary Fig. 2. Source: ABS [14–16] Ages 80 and above contributed to the increase in life expectancy in 2020 mostly due to lower mortality involving respiratory infections without COVID-19 (Fig. 3; Additional File 2, Supplementary Fig. 2). For mortality involving neither COVID-19 nor respiratory infections, ages 0–79 years was a larger contributor than 80 + years to life expectancy increase in 2020, while 80 + years was a larger contributor to life expectancy decline in 2021. In 2022, higher COVID-19 mortality contributed to each age group proportionally to overall life expectancy decline. Contribution of multiple and underlying causes Declines in mortality due to a range of underlying causes contributed to life expectancy increase in 2020, the largest being cardiovascular diseases (approximately 40% of total life expectancy increase). Over half the decline in 2022 was due to higher COVID-19 mortality as an underlying cause (Fig. 4). In 2023, approximately 40% of life expectancy increase was due to reductions in COVID-19 mortality, with cardiovascular diseases the next largest underlying cause contributor (approximately 25%). Fig. 4Contribution of underlying causes of death to change in life expectancy, Australia, 2017-19 to 2023. CVD: Cardiovascular diseases. Chron. resp.: Chronic respiratory diseases. Resp. inf.: Respiratory infections. NCD: Non-communicable diseases. Subst.: Substance use disorders. Other inf.: Other infectious diseases. Source: ABS (14-16) Contribution of underlying causes of death to change in life expectancy, Australia, 2017-19 to 2023. CVD: Cardiovascular diseases. Chron. resp.: Chronic respiratory diseases. Resp. inf.: Respiratory infections. NCD: Non-communicable diseases. Subst.: Substance use disorders. Other inf.: Other infectious diseases. Source: ABS (14-16) The lower deaths involving respiratory infections in 2020 had a wide range of underlying causes of death, including respiratory infections itself, cardiovascular diseases and cancers, none of which contributed more than one-quarter (Fig. 5, panel (a)). The sharp change in life expectancy due to deaths involving COVID-19 in 2022 and 2023 almost entirely had COVID-19 as the underlying cause (Fig. 5, panel (b)). For deaths involving neither COVID-19 nor a respiratory infection, the contribution to life expectancy increase in 2020 and 2023 had cardiovascular diseases as the most common underlying cause (Fig. 5, panel (c)). The negative contribution of deaths without COVID-19 nor a respiratory infection to life expectancy in 2021 occurred with the grouping of “other non-communicable diseases, other infections and substance-related disorders” being the most common underlying cause, with dementia also prominent for females. Fig. 5Contribution of underlying causes to change in life expectancy for (a) deaths involving respiratory infections (b) deaths involving COVID-19 and (c) deaths not involving respiratory infections nor COVID-19 (MCOD), Australia, 2017-19 to 2023. CVD: Cardiovascular diseases. Chron. resp.: Chronic respiratory diseases. Resp. inf.: Respiratory infections. NCD: Non-communicable diseases. Subst.: Substance use disorders. Other inf.: Other infectious diseases. LE: Life expectancy. Source: ABS [14–16] Contribution of underlying causes to change in life expectancy for (a) deaths involving respiratory infections (b) deaths involving COVID-19 and (c) deaths not involving respiratory infections nor COVID-19 (MCOD), Australia, 2017-19 to 2023. CVD: Cardiovascular diseases. Chron. resp.: Chronic respiratory diseases. Resp. inf.: Respiratory infections. NCD: Non-communicable diseases. Subst.: Substance use disorders. Other inf.: Other infectious diseases. LE: Life expectancy. Source: ABS [14–16] Life expectancy trends and contribution of MCODs by capital city classification All four regions in the capital city classification experienced a rise in life expectancy in 2020 compared with 2017-19, with “other capital cities” having the largest increase for females but the smallest for males (Fig. 6, Additional File 2, Supplementary Table 2). In Melbourne, despite having the highest number of reported COVID-19 deaths in 2020, life expectancy increase from 2017 to 19 to 2020 was the third-highest of the four regions. Life expectancy decline from 2020 to 2021 was the largest for males in Melbourne and females in other capital cities, while life expectancy in Sydney rose again for females. In 2022, all regions experienced declines in life expectancy, with Sydney having the largest falls for both males and females (almost 1.2 years for females and 0.8 years for males). However, in 2023 Sydney had the largest increase in life expectancy for each sex (almost 1 year), which was a year when life expectancy in all regions rose. In 2020-23 compared with 2017-19, life expectancy in Melbourne was almost unchanged, in contrast to other regions where it increased. The largest increase for males was outside capital cities and for females in capital cities other than Sydney and Melbourne. Fig. 6Life expectancy trends by capital city classification, Australia, 2017-19 to 2023. Source: ABS [14–16] Life expectancy trends by capital city classification, Australia, 2017-19 to 2023. Source: ABS [14–16] The contribution to the increase in life expectancy in 2020 due to lower respiratory infection mortality occurred everywhere except among Sydney females and was largest in females outside Melbourne and Sydney (Fig. 7). However, COVID-19 only had a small negative impact in Melbourne, which was more than offset by the positive contribution from other causes. In 2021, higher mortality not involving COVID-19 nor respiratory infections was the primary contributor to reduced life expectancy in each region, except Sydney where the rise in COVID-19 mortality had a bigger impact than in other locations. The large decline in life expectancy in Sydney in 2022 was almost all due to COVID-19 mortality for males but just over half from COVID-19 for females. COVID-19 was responsible for the entire or majority of declines in other regions. For 2023, the life expectancy rise in Sydney and Melbourne was more greatly contributed by declines in COVID-19 than elsewhere. Fig. 7Contribution of MCODs to change in life expectancy, by capital city classification, Australia, 2017-19 to 2023. Resp. inf.: Respiratory infections. Source: ABS [14–16] Contribution of MCODs to change in life expectancy, by capital city classification, Australia, 2017-19 to 2023. Resp. inf.: Respiratory infections. Source: ABS [14–16] Discussion Australia’s life expectancy experienced large fluctuations during the COVID-19 pandemic, with both females and males experiencing a significant increase in 2020 and a sharp fall in 2022 that was offset by a large rise in 2023. These annual changes were among the largest Australia has experienced in recent decades [20]. Overall, life expectancy in the 2020-23 period was higher than in 2017-19. This trend is in sharp contrast to other countries where life expectancy remained below 2017-19 for each year from 2020 to 2023, such as the USA, England and Wales and Canada [20]. Although Australia’s life expectancy trajectory differed from most countries during the pandemic, other countries which experienced similar fluctuations were New Zealand, South Korea, Norway and Finland males [20]. The large fluctuations in life expectancy reflect the relatively stringent lockdowns introduced in Australia in March 2020 and the after-effects following their relaxation in early 2022. Just over half the rise in life expectancy in 2020 was due to lower mortality involving respiratory infections such as pneumonia and influenza but not involving COVID-19, with this having a disproportionately large impact on people aged 80 years and above. This corresponds with findings that the restrictions reduced circulation of pneumonia and influenza compared with previous years and that in previous bad flu seasons older age groups have been particularly affected by these infections due to their frailty [7, 8]. Notably, the underlying causes of the lower respiratory infection deaths varied, including major non-communicable diseases of cardiovascular diseases and cancers, with only one-fifth having respiratory infections as the underlying cause. This demonstrates how MCOD data have exceptional value in settings with large fluctuations in mortality involving respiratory infections that are commonly an immediate cause of death. In Spain in 2020, when life expectancy fell sharply due to a high number of COVID-19 deaths, there was a sharp increase in pneumonia and influenza deaths measured as a contributing cause (i.e. MCOD but not underlying cause), but a decline when measured as an underlying cause [21]. The remainder of the increase in life expectancy in 2020 was due to deaths not involving respiratory infections nor COVID-19 (over 40%), which had a much younger age contribution and for which the most common underlying cause was cardiovascular diseases. Notably, injuries had a negligible impact as the underlying cause, given that pandemic-related restrictions may have been expected to lead to reduced traffic accident and workplace deaths. The decline in cardiovascular and other non-communicable disease mortality in 2020 that did not involve respiratory infections may have been due to continued decline in mortality from non-communicable diseases from before the pandemic. In 2021, when restrictions were still in effect in Australia, the decline in life expectancy had an older age profile and was mostly not due to COVID-19. This may have been due to people who were particularly ill or frail and whose deaths were averted in 2020 due to the restrictions then subsequently dying the following year. This phenomenon of the delay of deaths has been described as the “reverse harvesting” effect, that is, the opposite of the widely reported harvesting effect whereby short-term increases in mortality from heatwaves or pandemics results in the deaths of frailer people occurring earlier than otherwise and death rates subsequently being lower than expected [22–24]. The sharp decline in life expectancy in 2022 after the cessation of restrictions was mostly due to the rapid spread of COVID-19 which resulted in COVID-19 deaths that were nine times higher than in either 2020 or 2021. As expected, these COVID-19 deaths disproportionately affected older age groups because of their susceptibility to dying from the virus. However, this decline in life expectancy was relatively short-lived and was almost equally offset by an increase in 2023 that was half due to lower mortality from COVID-19 which had the largest impact on older age groups, as found in New Zealand, who had the sharpest rise in COVID-19 deaths in 2022 [25]. The direction of annual life expectancy trends was similar in each geographic region, despite the different stringency of restrictions experienced, but they differed in their magnitude. In Melbourne, which had the most stringent restrictions in 2020, life expectancy increased that year, especially for males; its relatively high COVID-19 deaths were more than offset by a strong decline in deaths involving respiratory infections, the spread of which was suppressed during its lockdowns. Melbourne’s relatively large fall in life expectancy in 2021 during its second year of stringent lockdowns was mostly due to a rise in deaths involving neither COVID-19 nor respiratory infections, which was likely an example of increased mortality of frailer people whose deaths were averted in this city’s first year of lockdowns in 2020 (see Additional File 1). Females outside Sydney and Melbourne, where restrictions were less stringent, had a large increase in life expectancy in 2020 due to fewer respiratory infection deaths but, like Melbourne, it was offset the following year by a fall in life expectancy due to a rise in deaths not involving COVID nor respiratory infections. In Sydney, where the lockdown was particularly strong in 2021, there was a rise in life expectancy for females that year which, like Melbourne in 2020, was due to higher COVID-19 deaths being more than offset by lower deaths from other causes. In contrast, life expectancy declined for Sydney males in 2021, possibly due to them experiencing a larger increase in life expectancy in 2020 than Sydney females, hence having a larger proportion of frail individuals in the population whose deaths were averted and susceptible to higher mortality in 2021. The relaxation of restrictions in 2022 that occurred in all of Australia resulted in a decline in life expectancy in each region because COVID-19 spread rapidly throughout the country. The largest life expectancy impact was found in Sydney, mostly due to a sharp rise in COVID-19 deaths and, for females, deaths not involving COVID-19 nor respiratory infections. In 2023, life expectancy increased for each region, with Sydney and Melbourne having the largest rise, which was especially due to lower COVID-19 mortality and which offset their lower life expectancy in 2022 for Sydney and 2021 and 2022 in Melbourne. Over the entire 2020-23 period, there were not major differences in life expectancy by capital city classification compared with 2017-19. This is attributable to the more stringent restrictions in Melbourne and Sydney in 2020-21 being effective at preventing further deaths from COVID-19 after local outbreaks and by all regions experiencing their biggest life expectancy falls occurring in 2022, primarily due to a sharp rise in COVID-19 deaths following relaxation of restrictions across the country, which was offset by a widespread abrupt recovery in 2023 [6]. Another possible cause is that COVID-19 vaccination rates at the commencement of the relaxation of restrictions at the end of 2021 were similar between these regions (Sydney 92.5%, Melbourne 91.2%, other capital cities 87.0%, outside capital cities 91.7%) [26]. A limitation of this analysis is the quality of reporting of causes of death on the death certificate. While MCOD data have several advantages, their quality is affected by how accurately the death certificate is completed, which may be affected by the physician’s knowledge of the disease history of the decedent and their recognition and knowledge of a disease’s role in leading to death. COVID-19 mortality reporting in death registration was relatively complete compared with COVID-19 mortality surveillance in the most populous state of New South Wales, with completeness being consistent by age and region of residence [27]. However the quality of reporting of other causes, including respiratory infections, could only be identified with closer examination of mortality data linked with other data sources. Furthermore, the quality of cause of death reporting could differ by place of death (i.e. hospital, home, aged care), but this variable was not available in our dataset. We also could not distinguish between community-acquired and hospital-acquired pneumonia, with the latter commonly having a bacterial rather than viral origin and is possibly affected by the number of hospital admissions. However, the rapid changes in mortality involving pneumonia following introduction and cessation of restrictions makes it highly likely due to changes in community-acquired pneumonia. Finally, more detailed analyses by age or cause were not possible because of ABS output rules, however these are based on numbers of cases less than 10 which would mean their overall impact on results would have been negligible. Conclusions This study demonstrates the high utility of MCOD data to understand the interaction of COVID-19, respiratory infections and non-communicable diseases during the COVID-19 pandemic and the role of pandemic-related restrictions, especially in a country like Australia where there was an initial life expectancy rise. MCOD data overcome certification guidance and coding rules that lead to the selection of the underlying cause of death but which mask the contributions of co-morbidities to death. This is especially relevant during the pandemic because COVID-19 is mostly reported as the underlying cause of death and the role of non-COVID-19 causes, especially respiratory infections like pneumonia, can be ignored by an underlying cause analysis. Furthermore, understanding the co-morbidities of COVID-19 is important given that the virus increased the risk of mortality for many co-morbid individuals [28]. Only very few MCOD studies have been undertaken of COVID-19, beyond relatively simple reporting by government agencies [29–31]. Underlying cause of death data are widely available from country statistical offices and international databases (e.g. Human Mortality Database, WHO Mortality Database) and hence enable standardised comparison across countries and over time by proving a single cause of death [3, 4, 11]. However, all countries that use the International Form of Medical Certificate of Cause of Death should have MCOD data and so should more routinely publish their results. Australia’s life expectancy trend was distinctive compared with other high-income countries during the pandemic. Government pandemic restrictions contributed not only to the large annual fluctuations in life expectancy but also that the geographic differences in these trends were small. These restrictions were particularly effective in 2020, reducing respiratory infection mortality as well as preventing COVID-19 deaths, as shown by the MCOD data. Such analysis of MCOD data provides a particularly nuanced understanding of mortality dynamics within a population during the pandemic, especially the interaction of COVID-19, respiratory infections and non-communicable diseases. Furthermore, MCOD analysis is also valuable outside of a pandemic especially given the increase in multi-morbidity at death due to an ageing population. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1. Supplementary Material 1. Supplementary Material 2. Supplementary Material 2.