The aim of this study was to assess COVID-19 impact on premature mortality using PYLL estimates from 17 countries and territories around the world from January to August 2020. Our analysis delivers three main key findings. First, the results evidenced COVID-19 as a cause of premature mortality in all the countries included in this study, with some countries being significantly more affected than others. Second, the largest proportion of PYLL was observed among the oldest (60+ or 65+) or the middle age group (40–59 or 45–64 years), depending on the country. Third, males died on average two years younger than females.
PYLL per person death
The estimates of the number of PYLL per death indicate that those dying due to COVID-19 in European countries, Israel and Australia were on average older than those who died due to COVID-19 in South America, Ukraine, Cyprus, Cape Verde, and Kazakhstan.
In our study, the average PYLL per person death was 8.7 years, which is lower than in other studies by several years. Arolas et al. (2020) estimated on average 14.5 years of life lost (YLL) per death in 42 countries [7]; Hanlon et al. (2021) computed 14 YLL in men and 12 YLL in women in UK per COVID-19 death [22]; and Elledge et al. (2020) counted 13.25 person-years death due to COVID-19 in the USA [23]. Mitra et al. (2020), who used a similar methodology to that used in this paper but for shorter time period (as of 30 May 2020), estimated 5.5 PYLL per person death in the USA [6] which is 2.4 years less than our estimate for USA (7.9 years).
The discrepancies between our results and some of the studies mentioned above, is likely explained by differences in methodology. Most of the above-mentioned studies used as the upper age limit the Japanese female’s life expectancy of 87 years as of 2019 (the highest life expectancy in the world) [24], WHO life tables or national life tables (which attributes YLL to all ages no matter the age of death) [6]. Using a higher life expectancy allows the inclusion of deaths in those older than 80 years in the YLL calculation, and thus leads to more YLL in countries with more deaths after the age of 80, while keeping constant the YLLs of countries with lower life expectancies (where most deaths would have happened before 80 years). This also explains why these publications used the YLL notation, rather than the more conservative PYLL term. Moreover, for the UK countries, where the PYLL per person death observed in this study were half of what was elsewhere reported [22], it was previously shown that accounting for prevalent comorbid conditions in the population when estimating PYLL, substantially reduces the PYLL estimates [22].
PYLL per 100,000 population
Europe showed the greatest differences in PYLL rates among the countries included in its geographical territory. Europe included some countries with a high number of PYLL per 100,000 such as England, Wales, and Scotland (numbers comparable to the PYLL in North America), as well as countries with very few PYLL per 100,000 population like Georgia, Slovenia, and Norway. This variability in PYLL among European countries follows the variable excess mortality experience of the European countries observed [9, 25,26,27,28,29,30].
Among South American countries, Peru had the greatest amount of PYLL per 100,000 population followed by Brazil and Colombia. A previous study, attributed the high burden of the COVID -19 disease in South America to the underlying social inequalities and to the public health limited capacities in the area [31].
In Europe, England & Wales, Sweden, and Scotland suffered more PYLL per 100,000 population than the rest of the European countries, supporting COVID-19’s impact on premature mortality. Interestingly, these countries were previously shown to have suffered excess mortality due to the COVID-19 pandemic, explaining in part the high number of PYLL per 100,000 population [9, 25,26,27,28,29,30].
In African, Asian, and Middle Eastern countries and in Australia, the PYLL per 100,000 population observed were relatively low. Cape Verde, the only African country participating in this study, experienced a different timing of the COVID-19 pandemic since they started seeing a surge in cases over the summer, but the peak of the pandemic was experienced in September–October 2020 [32]. Possible under reporting [33], seasonality patterns, possible pre-existing immunity to the virus [34], and the relatively young population may all have contributed to this low PYLL estimate [35]. To explain the low PYLL estimates in the Asian and Middle Eastern countries, similar hypotheses have been studied, including the environmental factor, possible resistance of East Asians to the coronavirus due to a gene mutation, cross-immunity, but also the early public-health measures taken by their authorities and the cautious behaviours taken by their populations [36, 37]. Lastly, Australia experienced a small peak in coronavirus cases in March, but the number of cases substantially escalated in July 2020, hence the low burden in terms of PYLL up to the end of August 2020 [38].
In July 2020, Oh et al. (2020), ranked the countries by the highest amount of PYLL per 100,000 population. They observed that Belgium, United Kingdom, Italy, Sweden, France and Spain had higher PYLL per 100,000 population than the USA and the other South American countries included in this study [18]. These results are not surprising, as by July, the European countries included in the Oh et al. study were some of the most affected countries in the world whereas, at the same time, in South America the pandemic had not reached its peak [39]. On the other hand, previously in May, Mitra et al. (2020) observed more PYLL in the USA than in Germany or Italy [6]. These results demonstrate that any comparisons between study results need to be interpreted with caution as the timing of observation of each investigation can heavily influence the results.
Our methodology (using an age limit of 80 years) also contributes to fewer PYLL than truly observed in countries with higher life expectancies (like most of European countries, Middle East, and Australia), resulting in the accumulation of fewer PYLL in these countries compared to countries with lower life expectancies (like the South American and African countries in this investigation).
Age-group differences in the number of deaths vs. PYLL
This study highlighted differences in the number of deaths versus the number of PYLL in the different age groups in the participating countries. As expected, the largest number of deaths was registered in the oldest age groups in all countries, whilst several countries observed the majority of PYLL in the middle age group (40–59 or 45–64 years). Similar results have been observed in other studies [6, 8]. Interestingly, in Peru and in Kazakhstan, most COVID-19 deaths happened in ages younger than 70 years. Possible incomplete data, underreporting in the elderly, lack of diagnostic test resources, and low proportion of elderly living in elder-care facilities (places with concentrate vulnerable people in conditions favourable to the spread of the virus) are potential explanations of this finding.
Across countries, a variable proportion of PYLL is observed for each age interval. At the same time there are large differences in the relative contribution of each age group to total PYLL, within countries which does not follow the sociodemographic index gradient observed in other studies [7].
Sex differences in the number of deaths and PYLL
According to the data used in this study, the largest number of deaths and PYLL happened in the same age groups in most countries, irrespectively of sex.
The comparison of the PYLL results by sex highlighted the increased burden of COVID-19 in terms of premature mortality in males rather than females. Looking at all countries together, 64.3% of the total PYLL were estimated among males and 35.7% among females. In all the countries except for Georgia, males suffered more PYLL per 100,000 deaths than females.
The greater burden of COVID-19 in terms of premature mortality in males rather than females, is also highlighted by the PYLL per person death results. On average males lost 9.5 PYLL per person death whereas females lost 7.4 years. The finding that COVID-19 premature mortality seems to be heightened in males than in females has also been observed in other studies [40, 41]. Biological reasons such as the immunological response between the two sexes could partly explain this difference [42, 43]. Channappanavar et al. (2017) demonstrated that oestrogen, the female sex hormone, can play a protective role by supressing the replication of SARS-CoV virus while activating the immune response [44]. Further, angiotensin-converting enzyme 2 (ACE2) is providing a protective role. ACE2 is the host receptor for SARS-CoV-2 virus, and it is also part of the renin-angiotensin system which is crucial in tissue response to viral infection [45]. The ACE2 is overexpressed in females, and increased levels of ACE2 is expected to provide greater tissue protection after viral entry, minimizing the death rates as compared to males [46]. However, other social factors such as the differential distribution of unhealthy behaviours, like smoking or alcohol consumption, as well as the increased likelihood of females to adhere to social precautions to reduce the exposure to SARS-CoV-2 might also contribute to these results [43, 47]. Lastly, given that females have a longer life expectancy than males, it is possible that the use of a common age limit may have led to an underestimation of PYLL among females. Nevertheless, further investigation in whether there is truly a gender gradient in COVID-19 premature mortality and the reasons behind it warrant further investigation [42, 48, 49].
PYLL per COVID-19 death definition
In this study, it was also observed that countries who reported as COVID -19 deaths, deaths where COVID-19 was either a cause of death or a contributing cause to death, observed on average more PYLL per 100,000 population than countries who adopted a stricter definition of COVID-19 deaths, and only reported deaths were COVID-19 was a cause of death. This result may be indicating that individuals where COVID-19 was present but perhaps not taking part in the chain of events leading to death, died on average at a younger age, pointing to the presence of comorbidities that most likely also explained the deaths. To this extent, this finding may be testifying to an over-estimate of COVID-19 deaths associated with the classification of all deaths in SARS-COV-2 positive persons as COVID-19 specific deaths. However, this estimate could be partly driven by the adoption of more inclusive definitions of a COVID-19 death by some of the most highly impacted countries in terms of excess mortality such as Brazil and the USA [21].
It may be worth highlighting here that in our study we report the PYLL estimates for each country based on the COVID-19 definition used at the national level. Regional comparisons are not expected to be significantly biased since almost all countries within a region follow the same definition.
PYLL rate against excess mortality during the study period
Lastly, of the countries displaying high PYLL rates in this study, all countries (except for Colombia and Israel) also experienced excess mortality per 100,000 population during the study period in 2020 compared to the mortality of the previous five years. This result suggests that excess mortality during the COVID-19 pandemic is accompanied by a large impact in terms of premature mortality.
Strengths and limitations
Our study has some important strengths compared to other published studies. It is one of the few studies attempting a comparison of the burden of COVID-19 premature mortality across 17 countries from different regions, some of which have not been previously studied. In addition, this study explored PYLL per 100,000 population and per person death and has also explored the differential impact of age and sex on these estimates. Furthermore, the data used in this study is primarily data from national sources, which is more reliable than publicly available data, often used in other similar publications.
This study has several data and methodological limitations. We obtained data up to the end of August 2020 (week 35) for all participating counties apart from Kazakhstan, where the data was available only until the end of week 31. As a result, we might have underestimated the PYLL for Kazakhstan compared to other countries, as well as the total PYLL estimate for the whole study period. Nevertheless, subgroup comparisons are not expected to be affected by this.
Another data limitation that poses a challenge in the comparison of results between countries is that each country grouped deaths in different age groups (see Supplementary Table S2). The estimates from countries with smaller range of age groups, would have more reliability than estimates from countries that have 20 or more years range age-groups, as in the latter case, the midpoint of the age group might be farther from the real age of each death.
Other studies found a strong association between the presence of comorbidities, such as cardiovascular disease, hypertension, diabetes, congestive heart failure, chronic kidney disease and cancer, and the risk of death due to COVID-19 [50]. The lack of accessible information on the presence of comorbidities among those who died in the countries included in these studies precluded us from correcting our results for comorbidities. However, aware of this limitation, we cautiously chose the methodology for this study, opting for a more conservative method than others have used. Other methodologies, like the national life tables or the WHO tables, that account for some few YLL no matter the age of the person, or the use of the Japanese female’s life expectancy as upper age limit, that is the highest in the world would lead to higher PYLL estimates which would not be as realistic given that the life expectancy of patients with the aforementioned comorbidities is shorter than that of the general population. Results of other studies where PYLL estimates are given before and after adjustment for comorbidities, support our choice in methodology [22].
The use of a standard life expectancy also has its limitations. The upper age limit was the same for all the countries and sexes, resulting in fewer PYLL in those countries with life expectancies higher than 80 years and in females, who on average die a few years later than men. However, despite its limitations, this methodology has been strongly recommended for comparison across countries [6, 51, 52].
Lastly, given the relevance of post-acute COVID-19 syndrome [53], it would be of much interest to include in investigations of COVID-19 burden, the metric of disability as in the Disability-Adjusted Life Year (DALY indicator). Even though such data was not available to the consortium at the time of publication, DALYs would provide more holistic estimates of health burden due to COVID-19 because the indicator takes into account post-acute and chronic effects which are likely to be more relevant among adults than premature deaths. Even though some attempts to estimate DALYs due to COVID-19 have been made [54, 55],more geographically diverse investigations are warranted.