Excess Mortality During the COVID-19 Pandemic in South Korea

: This study examines excess mortality in South Korea during the COVID-19 pandemic. I analyze age-speci  c mortality rates and present an international comparison. First, Korean excess mortality remained low until the end of 2021 but signi  cantly increased in early 2022. Second, this excess mortality was concentrated among older people. For example, cumulative excess mortality among the population aged 85+ years until the 30th week of 2022 was approximately 1-2 percent, that is, an additional 1-2 percent of this age group died compared with what we would have expected in the absence of COVID-19. Third, the international comparison demonstrates that excess mortality in South Korea was relatively low. The country experienced one of the lowest excess mortality rates among countries under study until the end of 2021, but excess mortality rapidly increased in early 2022. However, it returned to being comparatively low by mid-2022. This comparison shows cross-national variation in excess mortality, which may be associated with policy responses and public health infrastructure. Finally, I discuss implications and opportunities for future research.


Introduction
This study examines excess mortality after the outbreak of COVID-19 in South Korea (hereafter, Korea).It focuses on age and sex differences in excess mortality in Korea and presents an international comparison.Scholars have widely recognized that fatality rates due to COVID-19 are highly age-dependent.In other words, the risk of dying when stricken with COVID-19 is much higher for and concentrated among older persons than younger ones.For example, among the 29,209 cumulative deaths due to COVID-19 in Korea as of November 1, 2022, 59 percent (17,269 deaths) were individuals aged 80 years and over (Kye 2022).Furthermore, COVID-19 costs lives indirectly as well as directly.For example, the impact of the pandemic on the public health system certainly hindered its normal operation, resulting in unnecessary deaths that otherwise would not have occurred.In addition, some individuals may recover from the disease, but the consequent damage may be suf ciently signi cant to result in their deaths at a later point in time.However, such deaths are not attributed to COVID-19, despite certainly being related to the virus.Hence, researchers should estimate the numbers of deaths directly and indirectly caused by COVID-19.One way of doing so is assessing the magnitude of excess deaths.This study analyzes Short-Term Mortality Fluctuations (STMF) data to examine the patterns of excess mortality during the pandemic in Korea.The STMF visualization toolkit available at https://mpidr.shinyapps.io/stmortality/can produce the same results reported in this study because it allows for inspecting excess mortality under baseline mortality scenarios including the ones used in this study.The main contribution of this study is a reorganization of country-speci c excess mortality patterns and their comparison with pre-pandemic mortality.This will show how severe the mortality crises due to the pandemic were in different countries in a comparative perspective.The remainder of the paper is structured as follows.Section 2 brie y reviews previous studies that examined excess death during the pandemic.Sections 3 and 4 explain the data and methods used to estimate excess mortality, respectively.Section 5 presents the results, and Section 6 summarizes the  ndings and discusses their implications.

2
Literature review COVID-19 has been a major topic in demographic and public health research due to its global impact on public health, the economy, and social relationships.The increase in mortality during and after the pandemic is among the most prevalent research topics in these  elds.Scholars have analyzed various death outcomes and applied different methods to assess excess mortality.
A 2022 report by the World Health Organization (WHO) analyzed excess death counts using monthly death counts in 2020-2021 by specifying baseline deaths under the assumption of linear changes in monthly deaths (Knutson et al. 2023;WHO 2022).Global excess deaths between 2020 and 2021 were estimated at 14.9 million, which was more than  ve times the number of deaths directly related to .This  nding suggests that the number of deaths indirectly caused by COVID-19 far exceeded those directly caused by .
A few studies have also analyzed changes in life expectancy after the outbreak of the pandemic.For example, Aburto et al. (2022) used STMF data to analyze changes in life expectancy in 29 countries in 2020 and demonstrated that life expectancy in many countries declined in 2020 compared to 2019, and even was lower than in 2015 in certain countries (e.g., Russia).Although the annual increase in life expectancy at birth between 2015 and 2019 was approximately 0.2-0.3years, it declined in many countries after the COVID-19 outbreak.In the United States, life expectancy at birth decreased by more than 1 year, and the decrease in life expectancy was concentrated among the older population.However, there is cross-national variations, and the reduction in life expectancy among individuals aged 60 years and under was largest in Eastern Europe.
Other studies have examined excess mortality by comparing observed and projected life expectancies (Islam et al. 2021;Schöley et al. 2022).While Aburto et al. (2022) examined the trend in mortality after the pandemic, these studies compared observed and predicted mortalities under the assumption that previous trends on mortality continued.These analyses thereby accounted for declining mortality trends in assessing the size of excess mortality.The results demonstrated that the loss of life expectancy in Russia and the United States in 2020 was greater than 2 years, but no such reduction was observed in New Zealand, Taiwan, Korea, or Norway (Islam et al. 2021).In Germany, Luy et al. (2021) analyzed changes in life expectancy by age and sex and illustrated that life expectancy at birth and at age 60 in Germany decreased by 0.2 years, with a greater reduction among men than women.The reduction in eastern Germany was greater than in western Germany.
Changes in life expectancy can be easily interpreted, but require  ne-grained mortality data to be used.The STMF data provide fairly wide age intervals (0-14, 15-64, 75-84, and 85+ years) compared with the data usually used to construct life tables.Thus, we need to convert the data into shorter age intervals to estimate life expectancy, likely entailing errors.Hence, several studies analyzed age-speci c mortality rates instead of constructing life tables.Examining COVID-19 deaths in Italy in 2020, Basellini and Camarda (2022) found that the degree of intergenerational co-residence, number of intensive care unit beds per capita, and delay in the outbreak of the epidemic in uenced the magnitude of related deaths.Furthermore, Nepomuceno et al. (2022) used the STMF to analyze excess mortality by applying various mortality baselines.The authors constructed a total of 12 mortality baselines, four time-trend models (speci c-average, speci c-trend, speci c-average with trend, and harmonic with trend), and three reference periods (2010-2019, 2015-19, and 2017-2019).Their results indicate that excess mortality in the United States, Eastern Europe, and Southern Europe was high; in contrast, it was low in New Zealand and Taiwan.
The international comparison of excess mortality illustrates that excess mortality was lower in Korea than in the majority of countries.Moreover, the loss of life expectancy in 2020 was statistically nonsigni cant in Korea, in contrast to most other countries (Islam et al. 2021).The excess mortality in Korea in 2020 was one of the lowest among countries examined, regardless of estimation method (Nepomuceno et al. 2022).As elsewhere, excess mortality patterns in Korea showed withincountry variation.While excess mortality in Korea was negative in the majority of weeks in 2020, that is, fewer deaths occurred than expected based on the previous trend, excess mortality among men eligible for Medicaid was positive.This  nding suggests that the impact of the pandemic on mortality was concentrated among the economically disadvantaged (Kim et al. 2022).Shin et al. (2021) demonstrate seasonal and regional variations in excess mortality.Daegu-Gyeongbuk province, in which a cluster outbreak during the  rst wave of pandemic occurred, exhibited higher excess mortality during the spring of 2020.
Based on previous studies, the current study examines excess mortality in Korea from a comparative perspective.It contributes to the literature in two ways.First, it extends the time span of analysis on excess mortality during the pandemic.Previous studies in Korea analyzed data through 2020, when excess mortality was negligible (Kim et al. 2022;Shin et al. 2021).The larger comparative studies published thus far also focus on excess mortality in 2020 (Islam et al. 2021;Nepomuceno et al. 2022).By using data until the 30th week of 2022, this study extends previous work.Second, the current study explores age and sex differences in excess mortality in Korea during the pandemic and compares them with those of other countries.As harm due to COVID-19 on health is age-dependent, it examines differences in the patterns of excess mortality by age and sex.

Data
The study uses the STMF data provided by the Human Mortality Data Base (https:// www.mortality.org/),which provides annualized weekly mortality rates by age (i.e., 0-14, 15-64, 65-74, 75-84, and 85+ years) and sex for 38 countries (Jdanow et al. 2021).Time coverage varies by country, but data between 2015 and 2022 are available for the majority of countries.I use data last updated on September 26, 2022, which covers information until the 37th week of 2022 for several countries (e.g., Chile, Denmark, Scotland, and the Netherlands).However, data is more limited for other countries (e.g., Taiwan and Russia).Thus, the study analyzes data until the 30th week of 2022, when the data for Korea and the majority of countries are available.
Using short-term mortality data (e.g., weekly or monthly) introduces two concerns.If the late registration of deaths is common, then the death rates for the most recent periods could be underestimated.In addition, short-term  uctuation in death may re ect other shocks beyond the interest of the study.For example, the spikes in death rates in Europe from late July to early August in 2020 were largely consequences of a heat wave, rather than COVID-19 (Wang et al. 2022).Despite these concerns, I opt to use weekly mortality data, because doing so can enable the precise monitoring of trends (Nepomuneno et al. 2022).Furthermore, to minimize the undue in uence of a particular period on the trend, the study uses cumulative excess mortality, which is de ned in the next section.

Method
Specifying baseline mortality rates is necessary for analyzing excess mortality.Many methods can be used to de ne excess mortality (Nepomuceno et al. 2022).
The present study applies two approaches.First, I use a speci c-average method.
Weekly mortality rates tend to  uctuate cyclically due to the seasonality of mortality, and the speci c-average method uses the average mortality rates of the reference periods as baselines.Second, I use a speci c-trend method, which speci es baselines under the assumption that mortality rates per week change in a linear manner.Equations ( 1) and ( 2 The speci c-average model estimates baseline age-speci c mortality rates (Base y,w,a ) using the weekly average during the reference period (2015-2019), while the speci c-trend model assumes a linear time trend of weekly mortality.If mortality declines over time, then age-speci c baseline mortality rates under the speci ctrend model also tend to decline and are lower than those under the speci caverage model, which assumes no declining trend.In contrast, baseline mortality rates for the entire population will exhibit the opposite pattern, because crude death rates increase due to population aging.As excess mortality is de ned as the difference between observed and baseline mortalities, low baseline mortality yields high excess mortality.Thus, the speci c-trend method is likely to be more accurate than the speci c-average model, because the former re ects mortality trends, whereas the latter does not.Nonetheless, the study presents the results of the speci c-average method to assess the degree to which excess mortality depends on the assumption of baseline mortality.This study uses 2015-2019 as the reference period for specifying baseline mortality.The analysis of excess mortality in Korea uses both baseline mortality rates, while the international comparison uses only the speci c-average method.
Excess mortality rates are de ned as the differences in weekly mortality rates between baseline and observation, which is expressed in Equation ( 3).Excess mortality rates can produce negative values if the observed weekly mortality rates were lower than the baseline.
Excess y,w,a = Obs y,w,a − Base y,w,a , where y is the year, w indicates the week, and a is the age.
Cumulative excess mortality rates are estimated as the sum of excess mortality until a given week w, which is expressed in Equation ( 4): Cum_ Excess y,w,a = ∑ Excess y,w,a /52, where y is the year, w indicates the week, and a is the age.
Cumulative excess mortality rates at week w of year y display the excess mortality rates accumulated up to this time point since the outbreak of COVID-19.This measure is useful for showing how excess mortality accumulated during the period under consideration.Cumulative excess mortality increases with the increase in excess mortality.Although short-term events may in uence excess mortality, which could signi cantly  uctuate and render the interpretation of trends dif cult, cumulative excess mortality tends to remain stable and easy to interpret.The study (1) (4) will compare cumulative excess mortality rates with 2019 mortality rates, which will elucidate the severity of the mortality crisis during the pandemic.

Korea
Figure 1 presents the age-speci c annualized weekly mortality rates per 100,000 individuals in Korea for 2015-2022.The study presents the results using different y-axes by age groups to clearly show the trend for each age group.The  gure shows downward trends in the weekly mortality for those aged 0-14 and 65 years and over among men and women from 2015 to 2019.This result re ects the decline in age-speci c mortality rates over time.The study found no notable trend for those aged 15-64 years.In contrast, the mortality rates for the 75-84 and 85+ age groups signi cantly  uctuated, with a declining trend.In general, mortality rates were higher at the end and at the beginning of the year, which may be related to the increases in mortality during the winter.Mortality rates for the entire population tended to increase, which re ects population aging.Mortality trends after the outbreak of COVID-19 (since 2020) did not exhibit any notable difference from the previous period until 2022; a surge in mortality was observed for early 2022 except for the 0-14 age group, in which the declining trend Fig. 1  Notes: The reference period is 2015-2019, and the time unit is weeks.Linear refers to the speci c-trend method, and 5-yr avg.refers to the speci c-average method.Source: author's calculation using STMF data continued.Mortality rates returned to previous levels from May onwards.Although the study observes increases in mortality for all age groups except for the 0-14 age group, the increase in mortality was more signi cant for the older population.For example, the mortality rate for men aged 85+ years when mortality reached its high point in the Spring of 2022, was greater than 25,000 per 100,000.If this level of mortality remained constant for one year, then 25 percent of individuals in this age group would have died within a year.The study also shows gender differences: although mortality for men increased more than that for women in the 15-64 and 65-74 age groups during the peak period, the increase for women was greater for those in the 85+ age group.
Figure 2 presents baseline mortality rates for men and women.The blue lines depict baseline mortality rates using the speci c-trend method (linear), and the red lines represent those using the speci c-average method ( 5-yr average).Fluctuations produced by the speci c-trend method are greater than those by the speci caverage method, and baseline mortality rates using the speci c-trend model exhibit a declining trend that re ects the reduction in age-speci c mortality rates between 2015 and 2019.An exception is the 15-64 age group.In addition, the speci ctrend model shows increasing baseline mortality for the entire population due to population aging.The level of baseline mortality was higher for men than women, but the patterns of changes and differences between the two baseline mortality rates were similar.
Figure 3 depicts the excess mortality rates from the  rst week of 2020 to the 30th week of 2022.Although the study does not observe any remarkable trend for those aged 0-14 years, the changes exhibited by the other age groups display notable patterns.First, the study  nds no clear trends until early 2022.Second, excess mortality increased in Spring 2022 and then returned to the previous level.Third, excess mortality using the speci c-average model is lower than that produced using the speci c-trend model for all age groups, because the baseline mortality rates were lower for the speci c-average model than those for the speci c-trend model.For the entire population, the opposite is true; excess mortality rates for the speci c-average model tend to be higher than those for the speci c-trend model, because crude death rates increased due to population aging.Moreover, gender differences in excess mortality differed by age group.Except for those aged 85+, excess mortality among men was greater than that among women.For the entire population, excess mortality was higher for women than men, because the older population is composed of more women.
Figure 4 presents cumulative excess mortality rates.As is evident, the trends during the pandemic greatly differed from the baseline mortality rates.When comparing the age patterns of excess mortality under the speci c-average model (5-year average), we  nd that cumulative excess mortality rates remained negative except for women aged 15-64 years.In other words, weekly mortality rates after the COVID-19 outbreak, on average, remained lower than the average weekly mortality rates between 2015 and 2019.Thus, we can conclude that no excessive death tolls occurred during the  rst two years of the pandemic if the declining mortality trend is not considered.The study observes that cumulative excess mortality increased Notes: The reference period is 2015-2019 and the time unit is weeks.Linear refers to the speci c-trend method, and 5-yr avg.refers to the speci c-average method.Source: author's calculation using STMF data Notes: The reference period is 2015-2019, and the time unit is weeks.Linear refers to the speci c-trend method, and 5-yr avg.refers to the speci c-average method.Source: author's calculation using STMF data in early 2022 using the speci c-average model, which con rms a signi cant excess mortality during this period.Nonetheless, this increase did not turn the cumulative excess mortality positive.Here, women in the 15-64 age group were an exception, as their cumulative excess mortality was positive and increased over time.Interestingly, the cumulative excess mortality rates of the total population tended to increase for men and women despite the negative excess mortality for each age group, except for women aged 15-64 years.Simply put, crude death rates during the pandemic were higher than those during the reference period (2015-2019) and tended to increase.This result likely re ects the trend of population aging in Korea in general, rather than an increasing mortality risk per se.
The study  nds contrasting patterns in the speci c-trend model (linear).For the youngest age group (0-14 years), cumulative excess mortality remained negative and close to zero.The patterns for those aged 15-64 years differed by gender.Although cumulative excess mortality for men remained negative until the end of 2021 and increased in early 2022, that for women continuously increased since the COVID-19 outbreak.The levels of excess mortality among those aged 65+ years were higher for men than women under the speci c-trend model.For the 65-74 age group, cumulative excess mortality increased over time for men and women.For those aged 75 and over, it remained low or negative until the end of 2021, which then sharply increased in early 2022.The magnitude of cumulative excess mortality was greatest for those aged 85+ years: more than 2,000 additional men per 100,000 died since the outbreak of COVID-19.For the entire population, cumulative excess mortality was higher under the speci c-average model than the speci c-trend model due to population aging.The speci c-trend model re ects the increase in crude death rates due to population aging, whereas the speci c-average model does not.
Table 1 compares cumulative excess mortality until the 30th week of 2022 with mortality rates in 2019.Doing so enables the assessment of the relative magnitude of excess mortality since the COVID-19 outbreak.For the speci c-trend model, there were 94 excess deaths per 100,000 men, accounting for 15.1 percent of the crude death rate for men in 2019.This  gure amounted to 18.6 percent for women, because their crude death rate was lower than that for men.If we apply the speci c-average model, then these  gures became higher (men: 30.6 percent; women: 35.8 percent).A total of 15-35 percent of excess death occurred during the pandemic.As we observed in the analysis of excess mortality, the majority of excess mortality occurred in early 2022, which indicates the severity of the mortality crisis during this period.
The magnitude of cumulative excess mortality differs by age and sex.For those aged 0-14 years, cumulative excess mortality rates were negative under both models, which suggests that mortality rates were lower than the baselines.Under the speci c-trend model, cumulative excess mortality rates reached approximately 2-29 percent of the mortality rates for 2019 by age.These  gures were negative under the speci c-average model, except for women aged 15-64 years.Compared with mortality rates in 2019, the largest cumulative excess mortality was observed for people aged 65-74 years (men: 26.5 percent; women: 28.8 percent).Although cumulative excess mortality rates were highest among those aged 85+ years, the relative magnitude was the greatest among those aged 65-74 years.The cumulative excess mortality rates under the speci c-average model were negative except for women aged 15-64 years, which indicates that weekly mortality rates during the pandemic were, on average, lower than the mean mortality rates between 2015 and 2019.

International comparison
COVID-19 is a global phenomenon, and international comparisons are necessary to understand excess mortality during the pandemic.Many factors, including as baseline mortality levels and public policies, in uenced the rate of excess mortality in each country.This study has compared the level of excess mortality by applying baseline mortality rates under the speci c-trend model.It now presents excess mortality and cumulative excess mortality for the entire population of selected countries.Figure 5  Figure 6 compares the cumulative excess mortality rates.We can observe several patterns.First, cumulative excess mortality rates in East European countries increased nearly linearly.For example, the cumulative excess mortality in Bulgaria began increasing in the early stage of the pandemic and continued to increase, reaching 968 per 100,000 by the 30th week of 2022, for the highest rate among the countries examined.In other words, nearly one person per 1,000 Bulgarians additionally died across the two and a half years of the pandemic.Second, cumulative excess mortality rates rapidly increased during the early stage of the pandemic, but eventually slowed down in the United Kingdom and the United States.Excess mortality rates in these countries increased earlier than those of Eastern European countries.Although the former slowed down, the latter continued to increase.This contrasting pattern may be related to policy responses and overall public health capacity.Policy responses in the United States and the United Kingdom at the early stage were known to be ineffective in restraining the spread of the virus and excess death.However, these countries possessed advanced medical technology and health infrastructure to cope with the pandemic, leading to a slowdown in excess mortality.In contrast, Eastern European countries lacked suf cient resources and were therefore very susceptible to the virus and continued to experience high mortality rates.Third, countries in Northern Europe and Oceania Fig. 5  successfully restricted excess death.In New Zealand, cumulative excess mortality by the 30th week of 2022 was even negative, which indicated that mortality rates during the pandemic were lower on average than those assumed under the speci ctrend model.Finally, cumulative excess mortality in Korea was also comparatively low.Among the countries whose data were available until the 30th week of 2022, cumulative excess mortality for Korea was the  fth lowest after New Zealand, Luxemburg, Denmark, and Norway.Until 2021, cumulative mortality in Korea was the lowest globally along with New Zealand, but increased signi cantly in early 2022 and reached 96 per 100,000 by the 30th week of 2022. Figure 7 displays the relative magnitude of cumulative excess mortality by the 30th week of 2022 compared with the mortality rates in 2019.The cumulative excess mortality rates in Bulgaria accounted for 60 percent of crude death rates in 2019, which suggests that more than 20 percent more deaths occurred annually during the two and a half years of the pandemic than otherwise would have.This  nding illustrates that Bulgaria experienced a massive mortality crisis during the pandemic.As previously discussed, cumulative excess mortality was high among Eastern European countries and low in Northern Europe and Oceania.Cumulative excess mortality in Korea was close to 20 percent of the mortality rate for 2019, which was relatively low among the countries under study.Notes: The reference period is 2015-2019, and the time unit is weeks.Source: author's calculation using STMF data approximately 1-2 percent more deaths occurred for the oldest age group during the pandemic.Third, an international comparison showed that excess mortality in Korea was relatively low.Korea exhibited one of the lowest excess mortality rates globally until the end of 2021, but excess mortality increased rapidly in early 2022.However, the country's cumulative excess mortality was still comparatively low internationally as of the 30th week of 2022.The present study is largely descriptive and demonstrates substantial variations in excess mortality during the COVID-19 pandemic.This variation is likely related to differences in population structure, public health systems, and policy responses during the pandemic.Future studies should examine the more exact determinants of excess mortality during the pandemic.However, this is methodologically challenging.In particular, assessing the causal impact of policy responses on excess mortality could be fairly dif cult, because the level of excess mortality due to COVID-19 likely drove policy responses.Nevertheless, data on policy responses has been collected (Hale et al. 2020(Hale et al. , 2021)), which opens up the possibility of examining these relationships.This aspect could be an interesting next step for the research on mortality caused by COVID-19.
) present the formulas for the two abovementioned methods(Nepomuceno et al. 2022: 282-283): Speci c-average model: Speci c-trend model: Base y,w,a = β w h + ε y,w,a , (w = 1, 2,..., n; h = y−r, y−r+1,..., y-1)where y indicates the year, w is the week, a indicates age, r is the number of reference periods in years, n is the total number of weeks in year y, and ε y,w,a are the residuals.

Fig. 6 :
Fig. 6:Cumulative excess mortality rates per 100,000 in selected countries, entire population

Fig
Fig. 6:Continuation displays excess mortality rates at the population level.Excess mortality rates in several countries, such as New Zealand, Norway, and Australia, were very low, but Eastern European countries, such as Bulgaria, the Czech Republic, and Lithuania, displayed high excess mortality rates.Excess mortality in