Significant impacts of the COVID-19 pandemic on race/ethnic differences in USA mortality

The COVID-19 pandemic triggered declines in life expectancy at birth around the world. The United States of America (USA) was hit particularly hard among high income countries. Early data from the USA showed that these losses varied greatly by race/ethnicity in 2020, with Hispanic and Black Americans suffering much larger losses in life expectancy compared to white people. We add to this research by examining trends in lifespan inequality, average years of life lost, and the contribution of specific causes of death and ages to race/ethnic life expectancy disparities in the USA from 2010 to 2020. We find that life expectancy in 2020 fell more for Hispanic and Black males (4.5 years and 3.6 years, respectively) compared to white males (1.5 years). These drops nearly eliminated the previous life expectancy advantage for the Hispanic compared to white population, while dramatically increasing the already large gap in life expectancy between Black and white people. While the drops in life expectancy for the Hispanic population were largely attributable to official COVID-19 deaths, Black Americans additionally saw increases in cardiovascular disease and 'deaths of despair' over this period. In 2020, lifespan inequality increased slightly for Hispanic and white populations, but decreased for Black people, reflecting the younger age pattern of COVID-19 deaths for Hispanic people. Overall, the mortality burden of the COVID-19 pandemic hit race/ethnic minorities particularly hard in the USA, underscoring the importance of the social determinants of health during a public health crisis.


Main text
The COVID-19 pandemic has taken an unprecedented toll on mortality around the world. Most high-income countries experienced life expectancy losses in 2020 (1)(2)(3)(4) , and many continued to experience declines in 2021 (5). The United States of America (USA) saw its largest drop in life expectancy (1.7 years for females and 2.1 for males) in recent history (1), with COVID-19 deaths accounting for most of the decline for both females and males (6). Early data showed uneven impacts of the pandemic by race/ethnicity in the USA. The Center for Disease Control (CDC) estimates that between 2019 and 2020, life expectancy decreased by 3 years for the Hispanic population and by 2.9 years for the non-Hispanic Black (henceforth: Black) population, compared to a 1.2-year decline for non-Hispanic white (henceforth: white) people (6). The decrease was largest among Hispanic males (3.7 years) followed by Black males (3.3 years), and smallest among white females (1.1 years). These findings are consistent with early studies projecting a disproportionate impact of the pandemic on life expectancy among race/ethnic minorities (7). It is important to monitor these disparities which reflect underlying inequalities that are often magnified during a public health crisis (8). Both direct deaths from COVID-19 infection (9) as well as indirect deaths from other causes likely disproportionately affected race/ethnic minorities during the pandemic because of the social and economic disadvantages of historically marginalized populations in the USA (6,10).
This study provides a comprehensive analysis of mortality changes across racial/ethnic groups in the USA before and during the first year of the pandemic. It contributes to the evidence by: 1) analyzing recent trends in life expectancy, average years of life lost, and lifespan inequality from 2010 to 2020 separately for Black, Hispanic, and white populations; 2) identifying the ages and causes of death driving recent changes, including deaths from cardiovascular diseases, respiratory diseases, infectious and parasitic diseases, "deaths of despair" (i.e., suicide-, drug-, and alcohol-related mortality), cancers, accidents, and COVID-19; and 3) comparing race/ethnic gaps in these outcomes before and during the pandemic.
In the years before the pandemic, life expectancy in the USA followed atypical trends of stagnation that have not been observed in most high-income countries (6,11,12). These trends have been marked by worsening working-age mortality due to increased drug-related causes of death (13)(14)(15)(16)(17), as well as increased deaths from cardiovascular disease at middle and later ages (18). Life expectancy is consistently higher for the white population relative to the Black population, although the gap between Black and white people narrowed from 5.7 years in 2000 to 3.8 years in 2010 (19) . This convergence is partly due to relative improvements in mortality from heart diseases, HIV/AIDS, accidents, and cancer (20,21). By contrast, the Hispanic population had higher life expectancy than the white population throughout the pre-pandemic period, attributable to lower mortality from cancer, cardiovascular diseases, diabetes, chronic respiratory diseases, perinatal conditions, as well as "deaths of despair". Early evidence suggests that the pandemic has widened the white-Black gap in life expectancy, while reducing the Hispanic advantage (6). However, less is known about which ages and causes of death drove these changes (10).
The COVID-19 pandemic has directly and indirectly affected multiple causes of death. For example, delays in treatment may have increased mortality from cancers (22), or avoidance of hospitals for fear of infection may have increased mortality from acute cardiovascular events (23). COVID-19 is also associated with elevated risk of cardiovascular events and diabetes in the months following infection (24,25). Crucially, the impact of these changes likely varies across race/ethnic groups, due to differences in socioeconomic resources, rates of health insurance and access to health care (7). Recent findings show that, while COVID-19 death rates were highest in the Hispanic population, Black people experienced exceptionally large increases in mortality from heart disease, diabetes, and external causes of death (10).
So far, research on the differential impact of the pandemic on mortality across race/ethnic groups has mainly relied on estimates of overall life expectancy (6,7,26,27) and standardized death rates (10). While life expectancy is a widely-used and important indicator for studying population health and mortality, it is an average measure which conceals population variability and inequality (28)(29)(30)(31). Lifespan inequality is an important measure because it captures a fundamental type of inequality: variation in length of life (32). Two populations that share the same life expectancy could experience differences in the variation around the timing of death.
For instance, a high lifespan inequality measure would suggest that deaths occur within a wider age range, while a low measure of lifespan inequality would suggest a narrower age range.
Hence, lifespan inequality, measured as the spread of ages at death in a population (e.g. standard deviation), reflects how predictable length of life is at the individual level, and it underlies how uneven mortality improvements are at the population level (32,33).
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)  Black Americans not only experience shorter life expectancy compared to Hispanic and white people, but also have less predictable lifespans, with higher lifespan inequality (34,35). The impact of the pandemic on USA lifespan inequality is currently unknown. Evidence from England & Wales shows that both lifespan inequality and life expectancy decreased during 2020 because mortality was concentrated at older ages (36). However, in the USA, life expectancy losses during the pandemic have been driven by increases in mortality both at older and working ages (1,5,37). Previous studies show that increased midlife death rates increase lifespan inequality (38)(39)(40).  (42,43), AYLL is comparable across populations and over time, and is not affected by population age structure (44). This indicator enables researchers to quantify the burden of specific causes of death in a given year in a comparable way between populations (45). Using life expectancy, lifespan inequality and AYLL, we comprehensively quantify the unequal impact of age-and cause-specific mortality before and during the first year of the pandemic across race/ethnic groups in the USA.

Life expectancy, lifespan inequality and years of life lost
Life expectancy for both females and males stagnated in the second decade of the 21st century for all race/ethnic groups, but particularly for white people (Figure 1 Average years of life lost (AYLL) below age 95 varied from 11.7y for Hispanic females to 24.8y for Black males, with little-to no improvements over this period. Graphically, AYLL is the total colored area above the survival (probability of surviving) curve and can be attributed to specific causes of death. Deaths from cardiovascular disease (CVD) accounted for the largest . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. followed by cancer (more than 20%). In 2019, "deaths of despair" accounted for 10+% of AYLL in males, with the biggest impact for White males.
In 2020, life expectancy decreased for all groups (Figure 1, panel C). Hispanic and Black males saw the largest drops (4.5y and 3.6y respectively), while their female counterparts experienced losses of almost 3y. White females and males lost 1.2 and 1.5y of life expectancy, respectively.
Lifespan inequality increased slightly for Hispanic and White populations, but decreased slightly for the Black population. Premature mortality at younger ages is translated into increased lifespan inequality, but increases in mortality at older ages may decrease lifespan inequality. Hence, the divergent patterns in lifespan inequality across groups suggest varied impacts of the pandemic by age. AYLL also increased substantially for all groups. The largest increase in AYLL in 2020 was found in Hispanic males (4.1y), where COVID-19 contributed the most among specific causes (almost 25% of life lost) ( Figure 1, panel B), followed by Black males (3.5y) where COVID-19 accounted for more than 10% of the total years lost.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 4, 2022. 6

Trends in life expectancy and lifespan inequality by age and causes of death
The stagnation of life expectancy in the U.S. from 2010-2019 reflected changing age-patterns of mortality in certain causes of death. Improvements in cancer mortality, predominantly in ages 50-90, contributed to increases in life expectancy across all groups in this period ( Figure   2). The biggest gains from cancer improvements were observed for Black females and males (0.42y and 0.69y), and white males (0.48y) (see supplementary figure S2). Improvements in cardiovascular mortality also contributed positively to life expectancy in 2010-2019, especially for females (0.24y for white, 0.40y for Black, and 0.47y for Hispanic females). However, this progress was reversed by increased mortality from "deaths of despair", with males in all race/ethnic groups losing a staggering half a year or more from "deaths of despair" from 2010-2019. Losses from "deaths of despair" were concentrated in younger ages, with 30-39 yearolds suffering the greatest life expectancy losses in all race/ethnic groups. While females also lost life expectancy from "deaths of despair" in 2010-2019, the magnitude was smaller compared to males (0.16y).
The pandemic induced changes in the patterns of mortality by cause of death ( Figure 2, red color). In 2020, small cancer improvements translated into negligible gains in life expectancy, while deaths from cardiovascular mortality, "deaths of despair", and residual mortality increased dramatically. These changes translated into significant overall losses of life expectancy. For example, "deaths of despair" contributed to life expectancy declines of 0.35y for white, 0.46y for Black, and 0.33y for Hispanic males from 2019 to 2020, on top of already persistently high levels of "deaths of despair" in the previous decade. As expected, the largest contributions to life expectancy losses were from deaths due to COVID- 19 Increased mortality at any age translates into life-expectancy losses. However, for lifespan inequality to increase, mortality deterioration needs to be concentrated in younger ages. This means that worsening mortality at older ages, above a threshold age usually very close to life . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 4, 2022. ; https://doi.org/10.1101/2022.04.04.22273385 doi: medRxiv preprint expectancy, may lead to decreased lifespan inequality (46)(47)(48). For example, lifespan inequality decreased in England & Wales in 2020 due to worsening death rates at older ages during the pandemic. Causes of death that explain increases or decreases in life expectancy, are often not the same driving decreases or increases in lifespan inequality (49,50). In the USA lifespan inequality increased for all groups between 2010 and 2019. Increases were largely driven by mortality from "deaths of despair" below age 50 and improvements in cardiovascular disease mortality at ages 70+ (see supplementary figure S3). In 2020, lifespan inequality continued to increase for white and Hispanic people, while the Black population experienced small declines. Increased mortality at older ages from COVID-19 contributed to decreasing lifespan inequality among Black and white people, but not enough to see overall declines in lifespan inequality among Blacks. Among Hispanic males COVID-19 deaths contributed to increasing lifespan inequality because they were concentrated at younger ages, increasing uncertainty in lifespan. "Deaths of despair" continued to increase lifespan inequality in 2020 for all groups.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  (negative contributions in Figure 3). This means that if Black males were to experience the same mortality rates of "deaths of despair" as white males, the gap would be even larger.
In 2020, over the course of the pandemic, the Hispanic-white life expectancy gap narrowed to 2.1y for females and 0.7y for males, a decline of 1.7y and 3.1y, respectively. Conversely, the white-Black gap widened to 4.7y for females and 7.1y for males, an increase of 1.6y and 2.1y, respectively. The decrease in the Hispanic-white gap was driven by the disproportionate impact of COVID-19 deaths at working and older ages among Hispanic people, and doesn't reflect improvements in white mortality (negative values in Figure 3). The white disadvantage accounting for the gap was evident at ages below 60 from "deaths of despair", and from respiratory and cancers at older ages. The increasing white-Black gap in 2020 was largely explained by increased COVID-19 mortality above age 40, together with cardiovascular mortality. Other non-COVID cause-specific contributions to the white-Black gap remained similar to previous years. These trends in mortality also affected lifespan inequality (see appendix figure S6), but to a lesser extent. In 2020, the Hispanic population had the lowest lifespan inequality and the Black population had the highest.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 4, 2022. One year of pandemic-associated mortality nearly eliminated the Hispanic life expectancy advantage that was consistent over the 2010-2019 period. The declining Hispanic mortality advantage was predominantly driven by a greater concentration of COVID-19 deaths among working-age Hispanic people compared to the white population, a finding that has also been documented elsewhere (6,10). This is bolstered by our finding that lifespan inequality increased for the Hispanic population, indicating greater variability in ages at death among . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 4, 2022. ; https://doi.org/10.1101/2022.04.04.22273385 doi: medRxiv preprint Hispanic people in the USA. These patterns can be attributed to higher likelihood of viral exposure (due to employment and housing) and lower access to health care (52). The high burden of COVID-19 mortality in the Hispanic population, especially at working ages, suggests that exposure-related social factors outweighed any protection from previously accumulated health advantages. A key question is whether the Hispanic population will recover its life expectancy advantage in the short, medium or long term despite its socioeconomic disadvantage relative to the white population. This phenomenon is known in demographic research as the "Hispanic paradox" (64). Given that the reduction in the Hispanic-white life expectancy gap was largely explained by direct deaths due to COVID-19, vaccination rollout and age-specific uptake may be most relevant in the short term. Recent vaccination data suggests that Hispanic people are slightly more likely than white people to have received at least one dose (64% vs. 62%) (53).
In 2020, the persistent white-Black gap in life expectancy widened. This is the result of larger increases in COVID-19 and CVD mortality among Black people at older ages compared to their white peers, a finding that is consistent with previous studies (10). Although some of the CVD deaths in this population may be attributable to more underreporting of COVID deaths for Black people, the increase is dramatic and warrants further investigation. Black people in the USA may have experienced a "syndemic" of COVID-19 and other chronic diseases (51).
Our results suggest that, distinct from the Hispanic population, Black people may have been more exposed not only to COVID-19, but also to the indirect consequences of the pandemic, including delayed or missed healthcare (10,27). In contrast to the Hispanic-white gap, the white-Black gap in life expectancy may remain wider than it was in 2019 in the short-tomedium term, due in part to persistently lower vaccination rates among Black people relative to both White and Hispanic people (57% with at least one dose according to the KFF report).
It is plausible that this gap may continue to widen in 2022 because of the proportional increase in the hospitalization of unvaccinated Black adults during the Omicron-period (since December 2021) compared to other race/ethnic populations in the USA (54). Any indirect impacts of the pandemic that led to increases in CVD mortality may also persist longer than acute COVID risks.
Somewhat paradoxically, lifespan inequality decreased slightly for Black people, because the increase in mortality was primarily at older ages. A similar result was documented in England & Wales for 2020 (36). Typically, life expectancy increases are accompanied by decreases in . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 4, 2022. ; https://doi.org/10.1101/2022.04.04.22273385 doi: medRxiv preprint lifespan inequality (28,30,31). This means that countries that have achieved higher levels of life expectancy benefit from lower lifespan inequality. However, recent evidence has challenged the uniformity of this association. For example, in Eastern and Central Europe, life expectancy and lifespan inequality moved independent from one another before the dissolution of the Soviet Union due to worsening mortality at younger ages related to alcohol consumption and improvements in mortality at older ages (55). Similarly, groups with lower socioeconomic status in Finland and Spain experienced life expectancy increases and widening lifespan inequalities simultaneously usually due to high midlife mortality (32,56) . Our results are further evidence of the potential decoupling of life expectancy and lifespan inequality in the context of the pandemic, pointing to the importance of the age-profile of mortality during a crisis for changes to the predictability of lifespan.
The impact of the pandemic may be particularly unequal in countries characterized by high social inequality (e.g., as measured by the standard Gini coefficient), such as the USA (51).
Evidence from England -a comparatively unequal high-income country -suggests that South There are some limitations to this study, related to the availability and quality of the data that were analyzed. Many deaths for which COVID-19 was the underlying cause may have been attributed to CVDs, and this misreporting may have disproportionately affected the Black population, driving the increase in CVD deaths in this group (10). Finally, because of data limitations, this study examines three race/ethnic groups, and does not consider other minorities including American Indians and Alaska natives, Asians, and other race/ethnic groups.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 4, 2022. ; https://doi.org/10.1101/2022.04.04.22273385 doi: medRxiv preprint Despite these limitations, this study offers several important lessons for scientists and policymakers. The pandemic has affected mortality unequally. COVID-19 mortality hit Black and Hispanic Americans especially hard, exacerbating existing white-Black disparities and nearly eliminating the Hispanic mortality advantage. Our findings of higher Hispanic mortality at working ages are consistent with previous work suggesting increased risk of COVID-19 exposure for this group due to working conditions, housing, and access to preventive information (27,52). The widening of the white-Black gap in life expectancy is particularly concerning because, as estimated by Wrigley-Field (8), the difference between Black and white mortality pre-pandemic was already larger than the overall impact of the COVID-19 pandemic on white mortality.
Overall, our findings confirm the wide-ranging impact of the COVID-19 crisis on racial/ethnic differences in mortality in the USA. While the SARS-CoV-2 virus itself does not discriminate, the social environment shapes risk of infection and death in ways that reflect historical inequalities. Future work should continue to examine the direct impact of COVID-19 but also the impact of pandemic social and economic disruptions on racial/ethnic differences in health.
Besides risk of increased socioeconomic disadvantage during the pandemic, Hispanic and Black people also suffered much higher levels of bereavement and loss from COVID-19 in their own families(59), contributing to significant trauma and stress that may have long term health effects. Our study provides the first comprehensive assessment of the impact of the COVID-19 pandemic on racial/ethnic gaps in mortality, to ultimately inform research and policy interventions targeting such inequalities.

Data
We use data from the publicly-available United States multiple cause of death files, from the National Vital Statistics System division of the National Center for Health Statistics is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 4, 2022. ; https://doi.org/10.1101/2022.04.04.22273385 doi: medRxiv preprint (CVD), respiratory diseases, infectious and parasitic diseases (IPD), "deaths of despair" (suicides, drug-and alcohol-related deaths), cancers, accidents (excluding those from "deaths of despair"), and COVID-19. While "deaths of despair" encompass three causes of death, driven by underlying phenomena that are likely independent (13,14), evidence suggests that all three causes were rising before and during the pandemic. As such, we combine them here.
Specific International Classification of Disease (ICD)-10 codes are listed in Table 1. Life expectancy at birth is a summary indicator of population health and mortality. It expresses the average number of years a synthetic cohort of newborns is expected to live if they were to experience the mortality rates observed in a given year (61). It is not affected by population size or age-structure, which makes it a preferred indicator for comparisons over time and between populations.
Average life years lost (AYLL) is a complementary indicator to life expectancy. It refers to the average years lost of a synthetic cohort due to mortality in specific age-groups or causes of death (41). For example, if individuals live on average 80 years between birth and age 95, then AYLL equals 10 years. In this analysis we set the upper limit of 95 to quantify AYLL. This strategy has been used previously to characterize mortality in contexts where subpopulations experience different mortality profiles, e.g., population with and without mental health disorders (44).
Lifespan inequality refers to variation in ages at death and is a marker of heterogeneity in mortality at the population level and reflects uncertainty in lifetimes at the individual level (32).
The greater lifespan inequality, the more unpredictable lifespans are in a population. We measure lifespan inequality with the standard deviation of the age-at-death distributions.
Several indicators exist to measure lifespan inequality including the life table entropy, the Gini coefficient, or life disparity, and all are highly correlated when measured from birth (62), which suggest that our results would not change on the basis of the indicator chosen.

Decomposition of life expectancy by age and cause of death
In order to disentangle age-and cause-specific effects over time and between groups for life expectancy and lifespan inequality we used the linear integral decomposition method (63), a state-of-the-art method that allows us to decompose the difference of two values of life expectancy or lifespan inequality by age and cause of death, which has been implemented previously for this type of analysis (e.g. ref. (1)).
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 4, 2022. ; https://doi.org/10.1101/2022.04.04.22273385 doi: medRxiv preprint Figure S4. Results as shown in Figure 3 for Hispanic-Black difference in life expectancy.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 4, 2022. ; https://doi.org/10.1101/2022.04.04.22273385 doi: medRxiv preprint . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 4, 2022. ; https://doi.org/10.1101/2022.04.04.22273385 doi: medRxiv preprint