As European countries are struggling to contain the third wave of the coronavirus disease 2019 (COVID-19) pandemic and limit the spread of the more infectious and deadlier new variants of the virus, governments face a difficult trade-off between supporting the economy and protecting public health. Therefore, it is crucial that officials understand the direct and indirect health effects of the pandemic when making policy decisions.

COVID-19-related mortality is one key indicator that is widely used to track the severity and the public health effects of the pandemic. When the pandemic began in early 2020, most of the existing literature on the impact of the pandemic relied on case-fatality rates (CFR) as a measure of mortality (CDC COVID-19 Response Team, 2020; Giangreco, 2020; Khafaie and Rahim, 2020; Onder et al., 2020). However, CFR are not informative for international and historical comparisons. Since they are calculated as the number of deaths divided by the number of confirmed cases, the absence of an accurate estimation of the infection rates in a reference population makes the denominator in the CFR reliant on testing strategies and capacities.

There is no uniform way of classifying, recording and reporting COVID-19 deaths (Garcia-Basteiro et al., 2020). Moreover, when the epidemic worsens, the counting of fatalities becomes more difficult. People who die at home or in long-term care facilities might not be tested at all simply because resource allocation prioritises emergency operations (Iacobucci, 2020; O'Dowd, 2020). Likewise, there may be indirect mortality effects due to congestion in healthcare services (The Lancet Oncology, 2020), or to patients with chronic conditions avoiding visiting health facilities because they are concerned about the risk of COVID-19 infection (Weinberger et al., 2020). Therefore, COVID-19 mortality reports that rely on data on COVID-19-attributed deaths are likely to undercount the pandemic's death toll.

With the release of mortality surveillance data, such as all-cause mortality data from vital statistics systems for various countries, recent studies have used the “excess deaths” approach to estimate the mortality burden of the COVID-19 pandemic (Rivera et al., 2020; Rossen, 2020; Stang et al., 2020; Vandoros, 2020). Excess mortality counts the total number of persons who have died, regardless of the cause of death, relative to the number of deaths that would have normally been expected for a given place and time. For instance, Modi et al. (2020) compared excess mortality data for Lombardy with the official fatality statistics for Italy, and found that the estimated excess mortality in Lombardy between January and April 2020 was about three times higher than the COVID-19 death rate reported in the official data. Thus, this measure allowed the authors to capture both under-reported COVID-19-related deaths and fatalities that could be indirectly attributed to a lack of health care access, economic deprivation or other causes.

While excess mortality is a useful measure of the health impact of the COVID-19 pandemic, using overall crude death rates or the proportion of deaths for cross-national or historical comparisons is not very informative, because these indicators are affected by the age distribution of the populations studied. By contrast, life expectancy, which is calculated based on human mortality data aggregated in life tables, is insensitive to the age structure of the population, and can therefore reflect differences in mortality reasonably well. Against this background, this study aims to measure the impact of the first wave of COVID-19 on life expectancy at birth by focusing on the hardest hit areas in Italy. As Italy was the first western country severely affected by a large COVID-19 outbreak, this approach allows us to reasonably capture the human cost of the first wave of the COVID-19 pandemic, especially in a context in which non-pharmaceutical interventions were delayed.

In particular, this study focuses on specific geographical areas in Italy that were the most severely affected by the early stages of the COVID-19 pandemic: four provinces in Lombardy (Bergamo, Lodi, Cremona, Brescia) and one province in Emilia Romagna (Piacenza). In modelling the spread of COVID-19 in Italy, Gatto et al. (2020) highlighted the importance of considering the spatial nature of the progress of the wave of infections. The selected provinces experienced the highest numbers of excess deaths in Italy in the observation period compared to the average mortality levels in the years 2015–2019. The highly clustered nature of local transmission resulted in a high concentration of severe illnesses and deaths in one area (Jia et al., 2020). Therefore, the direct impact of COVID-19 on mortality and average life expectancy was likely felt at the sub-national level, rather than at the national level. Indeed, our results suggest that even in Lombardy – which was the hardest hit region in Italy during the first wave of the pandemic (Sebastiani et al., 2020) – the reduction in life expectancy due to COVID-19 was significantly lower than in Bergamo, the province that contributed one-third of the total excess mortality in the Lombardy region. Thus, spatial granularity is needed to assess the full scale of the impact of the pandemic on human life.

Drawing on daily death registration data published by the Italian Statistical Office (ISTAT) for the period of 1 January to 31 October 2020, the present study compares the mortality rates in 2015–2019 and in 2020 across age and gender categories, and provides estimates of the changes in life expectancy following the first wave of the COVID-19 pandemic. While measures such as mortality rates are no doubt useful, they need to be collapsed in an index that is universal enough to provide a reliable measure of all of the human lives lost. By contrast, life expectancy is significantly related to the overall wellbeing of the population, and can therefore provide a simple, objective and immediate measure of the human casualties associated with unprecedented shocks, such as the COVID-19 pandemic (Aburto et al., 2020; Ghislandi et al., 2019; Sen, 1998). Furthermore, as reliable measures of life expectancy are available for some countries from the 19th century onwards, we can use life expectancy for historical comparisons of the human costs associated with major events.

In the early hours of 21 February 2020, the first severe case of local transmission of COVID-19 was diagnosed in Europe at a small hospital in Codogno, a municipality in the province of Lodi, south-east of Milan (Paterlini, 2020). Initially, authorities reacted by tracing the connections of patient one, but ultimately failed to identify a patient zero. As early as 24 February 2020, 11 municipalities in the province of Lodi were placed under strict measures to contain the spread of the disease, and were declared a quarantine “red zone”. Meanwhile, another cluster of COVID-19 cases emerged in Alzano Lombardo and Nembro, two municipalities in the province of Bergamo, north-east of Milan. In response to the rapid rise in the number of detected cases, especially in the municipalities surrounding these two epicentres, the Italian government announced on 8 March 2020 that it was imposing a (partial) nationwide lockdown starting on 9 March, followed by a total lockdown of all non-essential activities starting on 23 March (Galizzi and Ghislandi, 2020). While the Italian government was praised by the World Health Organization (WHO) for implementing such drastic measures (i.e., restrictions that had not been employed in modern democratic nations since World War II), the virus had already been spreading undetected in the northern part of the country since December 2019 (La Rosa et al., 2021). Thus, it appears that these containment measures were imposed a little too late (Signorelli et al., 2020). During this first wave of the pandemic, the outbreak put an unprecedented burden on the Italian healthcare system, resulting in an exceptionally high number of coronavirus deaths.

Geographically, Lodi and Codogno – two of the 12 provinces in Lombardy – are close to the other two provinces included in our sample: Cremona and Piacenza (see Figures A.1 and A.2 in the Appendix for the geographical location of the provinces being studied). The epidemic wave involving these provinces is thus considered as part of the Lodi-Codogno cluster. Bergamo and Brescia are located north-east of Milan, and, even though the first severe cases of COVID-19 were detected in these provinces just one day after patient one was identified in Lodi, they experienced a week-long delay in the arrival of the first epidemic wave (Galizzi and Ghislandi, 2020).

Of the regions in Italy, Lombardy is the most populated, and it has the highest Gross Domestic Product (GDP). Overall, one-sixth of the Italian population live in Lombardy, and the region produces one-fifth of the country's GDP. Lombardy is relevant for our analysis, because it was the region in Italy that was hardest hit by the COVID-19 pandemic during the first wave, accounting for almost 50% of the human casualties in the entire country (Odone et al., 2020). Indeed, with the exception of Piacenza (located in the Emilia Romagna region), a province that borders the Lombardy region, all of the other four hardest hit provinces included in the analysis are located in Lombardy. Thus, in the following, we will also present statistics for the region of Lombardy.

Figures 1(a)–1(f) show the trends in daily mortality for the five provinces with the highest numbers of declared cases in Italy and in the whole Lombardy region.² Plotting the mortality distribution by age groups allows us to fully capture the progression of the first epidemic wave. It is evident that the epidemic curve inflated with age across all provinces. It is also clear that by 30 April, the daily mortality in all selected provinces approached the pre-pandemic values (i.e., no excess mortality). Hence, the wavelength of the epidemic in these provinces was between six and eight weeks, with the peak happening around two weeks after the onset of the outbreak.

The vertical lines show four relevant dates for the evolution of the first epidemic wave. After the case of patient one was first identified in Codogno, located in the province of Lodi, the authorities quickly locked down 11 municipalities in the area on 24 February 2020. The containment measures associated with the lockdown were not implemented in other provinces until after 8 March. Although the earlier lockdown enabled Lodi to flatten the curve more effectively than other severely affected provinces (Figure 1), the province still experienced a notable increase in excess mortality. Considering that the incubation period – i.e., the time between the exposure and the onset of symptoms – can be up to 24 days, it is evident that the lockdown was imposed too late in these provinces. While political reasons prevented the authorities from implementing the lockdown earlier in the provinces where the number of cases had been rising rapidly, like in Bergamo, there is recent evidence showing that COVID-19 had already been circulating undetected in northern Italy since December 2019 (La Rosa et al., 2021). Thus, our study proxies the impact of the COVID-19 outbreak in the absence of containment interventions.

The geographical distribution of excess deaths in the first quadrimester across Italy (Figure 2) matches the distribution of confirmed cases (which comprise the deceased, the recovered individuals and the active cases) provided by the Italian Civil Protection Department, which publishes the official surveillance data on COVID-19.³ This geographical pattern indicates that the excess mortality observed in our data represents mortality directly and indirectly related to COVID-19.

Note that in Figure 2, we focus on the 1 January-30 April period only in order to better capture the impact of the first wave of the COVID-19 pandemic. Compared to the average number of people who died in the same period in the previous five years (2015–2019), the excess number of deaths (for those aged 40 or older) between 1 January and 30 April 2020 sums to 6,084 in Bergamo, 3,969 in Brescia, 2,030 in Cremona, 905 in Lodi and 1,170 in Piacenza. For the entire region of Lombardy, the excess number of deaths is approximately 23,649 (Table 1). The total number of COVID-19 deaths reported by the Italian Civil Protection Department for Lombardy as of 30 April 2020 is 13,772. This implies that the overall death toll of the first epidemic wave was about 70% higher than that suggested by official statistics on COVID-19 deaths. The mortality rate in the first quadrimester of 2020 increased substantially in all provinces and for all age groups, with the largest increase being observed for men aged 70–79 in Bergamo (a 347% increase). Age clearly represented a risk factor for excess mortality, in line with the age gradient in COVID-19 CFR observed in Italy and elsewhere. For instance, among the excess deaths observed in Bergamo, the mortality rate was much higher among older men aged ≥70 years. A similar ratio is found in the other provinces.

When we only consider the distribution of excess mortality without adjusting for population size in each age-sex category, we observe slightly more excess mortality in men than in women (53% of excess deaths involved male subjects). However, when we consider the mortality risk ratio between the sexes, we find that the excess mortality for males was consistently higher than that for females across all age groups and provinces (relative risk ≥1).

The trends in the first quadrimester and the annual life expectancies are illustrated in Figures 3 and 4. When we look at the trends in the first quadrimester of 2020, it is evident that the drop in life expectancy was significant for both men and women in all provinces. Compared to the average life expectancy of the 2015– 2019 period, the reduction for men ranged from 5.5 years in Brescia to 8.1 years in Bergamo, and the reduction for women ranged from 4.1 years in Piacenza to 5.8 years in Bergamo. The larger reduction in the first quadrimester life expectancy for men was due to sex differentials in the COVID-19 mortality risk, as both the official case fatality data and our death registration data consistently show. Indeed, when we decompose the loss in life expectancy to identify which age groups contributed the most to the reduction in life expectancy (Figure A.3 in Appendix), it becomes clear that the older populations, and especially men aged 60–79 years, played a major role.

When life expectancy is extrapolated for the whole year, the loss in life expectancy is diluted over a longer period. Thus, the drop in life expectancy due to COVID-19-related excess mortality was less steep than that observed in the first quadrimester life expectancy.

In the most severely hit province of Bergamo, life expectancy dropped by 4.1 years for men and 2.8 years for women when compared to life expectancy for the years 2015–2019. In the slightly less affected provinces of Brescia, Cremona, Lodi and Piacenza, the reduction in life expectancy ranged between 2.4 in Brescia and 3.8 in Cremona for men, and between 1.9 in Piacenza and 2.6 in Cremona for women. As expected, the reduction in life expectancy was smaller in Lombardy, at 1.9 years for males and 1.5 years for females.

When we turn to the national level, we see that the results are extremely heterogeneous (Figure 5). It is evident that the higher excess mortality experienced in the northern part of Italy, particularly in Lombardy, was not experienced in most of the provinces of the central part and the south of the country. For example, in provinces like Sassari and Nuoro in Sardegna and Cosenza in Calabria, the lockdown reduced mortality in the first four months of the year, resulting in an estimated increase in life expectancy up to two years for both men and women.

By avoiding the inconsistencies in the classification of causes of death and in testing practices, and by focusing on the five areas in Italy that were most severely affected by the first wave of the pandemic, this study provided an assessment of the full impact of the first wave of the COVID-19 pandemic on human life.

Two empirical regularities clearly emerged when we looked at demographic differentials. First, the age gradient in excess mortality was steep, and age was the most evident risk factor for COVID-19 mortality. In Lombardy, men and women over age 70 were 23 times more likely to die than their counterparts under age 70. These patterns were replicated in all five provinces. Therefore, areas where older people made up a high proportion of the population (e.g., 17% of the population were over age 70 in the Lombardy region in 2019) had a higher burden of COVID-19 mortality (Dowd et al., 2020). Second, within each province, the risk of dying was consistently higher for men than for women for all age classes and provinces considered. Evidence that men are more likely than women to suffer from COVID-19, as measured by hospitalisations, admissions to intensive care units and fatality rates, has been consistently reported for other countries across different studies and subsamples (Gebhard et al., 2020; Peckham et al., 2020; Scully et al., 2020). Higher mortality rates for men than for women translate into a larger reduction in life expectancy for men than for women.

Although these data provided evidence of the severity of the first wave of the COVID-19 pandemic in Europe, a further measurement effort was needed, particularly for geographical and historical comparability purposes. In terms of life expectancy, we showed that for the period of 1 January to 30 April 2020, the reduction in the first quadrimester life expectancy, compared with the average of the years 2015–2019, was as high as 8.1 years for men and 5.8 years for women in Bergamo.

When the analysis was extended to the whole year, under the assumption that the mortality rates from November onwards were back to “normal”, life expectancy was reduced by up to four years (for men in Bergamo). However, significant uncertainties remain about the longer-term effects of the pandemic on health conditions among, for instance, patients who recovered from COVID-19 with major co-morbidities and mental health issues, and pregnant women. It is also possible that indirect physical and mental health consequences of changing socio-economic conditions affected the mortality patterns in 2020.

What can we say regarding the validity of the no harvesting assumption? Figure 1 provides evidence that after the end of the first wave, the mortality patterns in all age-provinces groups were largely similar to those in the previous years. This result is not consistent with harvesting, which would require negative excess mortality to compensate for the high levels of mortality registered in the first quadrimester. It should also be noted that since November 2020, Italy has been experiencing a severe second wave of infections that has not fully finished. Moreover, since March 2021, the country has been bracing for a third wave. Therefore, mortality in Italy is likely to increase even further. Thus, the figures provided can be considered estimates of the human life lost only for the first wave of the COVID-19 epidemic in the affected provinces.

It should also be noted that in the first quadrimester, some provinces in Italy experienced an improvement in life expectancy thanks to a reduction in mortality compared to the average of the previous years. There may have been spill-over benefits of the lockdown measures that contributed to a decline in premature deaths, such as from road traffic fatalities, alcohol consumption, violence and injuries at work (Qi et al., 2020; Qureshi et al., 2020). Moreover, our observation that the epidemic had a differential impact across different regions within Italy shows the importance of considering specific geographic areas when estimating the effect of the COVID-19 pandemic on human life. Indeed, focusing on national-level statistics only would further bias downward the estimation of the impact of the virus. Thus, our explicit focus on a local context can be considered the main strength of this analysis. Because the COVID-19 outbreaks have been geographically concentrated, looking at country-level life expectancy is misleading, and underestimates the actual impact of the pandemic.

Along with Italy, other European countries have been experiencing sharp declines in life expectancy due to the COVID-19 pandemic. Estimating weekly life expectancy for Spain, Trias-Llimós et al. (2020) found a particularly large drop at the beginning of April 2020, with a decline of up to 7.6 years at the national level. At the regional level, the authors reported an even more pronounced drop in life expectancy, with Madrid in particular experiencing a large reduction, ranging from 11.2 years in week 13 to 14.8 years in week 14 for both men and women. Moreover, the findings of a study for Sweden based on mortality data for the first 33 weeks of the pandemic suggest that life expectancy at age 50 in Stockholm decreased by about two years for men and about 1.5 years for women (Modig et al., 2021). At the national level, the reduction in annual life expectancy during the COVID-19 pandemic is expected to be smaller. Similarly, a study that calculated life expectancy at birth in England and Wales on the basis of data for the first 47 weeks of the pandemic found that it declined by 0.9 years for women and 1.2 years for men between 2019 and 2020 (Aburto et al., 2021). Our results are in line with those of these previous studies, as we also found the largest declines in life expectancy at a local level; in our case, in the north of Italy.

Under normal conditions, life expectancy at birth is calculated with mortality data for one calendar year, and provides an estimate of mean longevity for a hypothetical group of individuals who experience the mortality regime of a given period over their entire life course. Obviously, in reality, no group of people will be exposed over their life course to the mortality regime of the worst hit regions in Italy during the first wave of the COVID-19 pandemic. With the development, approval and rollout of several vaccines and the implementation of protective measures, it is highly likely that in the future, mortality in these regions will bounce back to lower levels. Still, life expectancy is a powerful tool for summarising and comparing mortality rates between regions and over time, especially because it accounts for differences in age-specific mortality (Marois et al., 2020; Trias-Llimós et al., 2020).

As the results of this study show, the cost in terms of human life of the delays in public interventions to reduce the transmission of the virus was disturbingly high. As European countries struggle to manage the successive waves of the coronavirus by striking a balance between protecting public health and reducing the economic effects of restriction measures, it is important to keep in mind the potential risk of viral reintroduction, and the direct and indirect dangers it poses to human life. Well-planned government measures aimed at flattening the epidemic curve while preventing a new wave of infections, along with public cooperation in maintaining physical distancing, wearing a face mask and practicing proper hygiene until there is widespread access to vaccination for the novel coronavirus, are key to achieving a balance between protecting public health and sustaining the economy.

SG and BS designed the research; BS acquired data, performed the analysis and created the display items with the help of MS. SG, RM, MS and BS contributed to the writing of the manuscript.