Artículo de revista: Pifarré i Arolas, Héctor; Acosta, Enrique, Dudel, Christian; Mhairi Hale, Jo; Myrskylä, Mikko (2023) ‘US Racial–Ethnic Mortality Gap Adjusted for Population Structure’. Epidemiology, 34. pp. 402-410 (ISSN 1044-3983) DOI: 10.1097/EDE.0000000000001595 Original Article US Racial–Ethnic Mortality Gap Adjusted for Population Structure Downloaded from http://journals.lww.com/epidem by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1 Héctor Pifarré i Arolas,a,b Enrique Acosta,c Christian Dudel,c,d Jo Mhairi Hale,c,e and Mikko Myrskyläc,f AWnYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC4/OAVpDDa8KKGKV0Ymy+78= on 04/25/2023 from circulatory diseases. The disadvantage is 72% among Blacks Background: US racial–ethnic mortality disparities are well doc- (men: 47%, women: 98%) and 65% among Native Americans (men: umented and central to debates on social inequalities in health. 45%, women: 92%), larger than life expectancy measured disad- Standard measures, such as life expectancy or years of life lost, are vantage. In contrast, estimated advantages for Asian Americans are based on synthetic populations and do not account for the real under- over three times (men: 176%, women: 283%) and, for Hispanics, two lying populations experiencing the inequalities. times (men: 123%; women: 190%) larger than those based on life Methods: We analyze US mortality disparities comparing Asian expectancy. Americans, Blacks, Hispanics, and Native Americans/Alaska Natives Conclusions: Mortality inequalities based on standard metrics’ syn- to Whites using 2019 CDC and NCHS data, using a novel approach thetic populations can differ markedly from estimates of the popula- that estimates the mortality gap, adjusted for population structure by tion structure-adjusted mortality gap. We demonstrate that standard accounting for real-population exposures. This measure is tailored metrics underestimate racial–ethnic disparities through disregard- for analyses where age structures are fundamental, not merely a con- ing actual population age structures. Exposure-corrected measures founder. We highlight the magnitude of inequalities by comparing of inequality may better inform health policies around allocation of the population structure-adjusted mortality gap against standard met- scarce resources. rics’ estimates of loss of life due to leading causes. Results: Based on the population structure-adjusted mortality gap, Keywords: Age structure; Mortality; Exposure; Racial disparities Black and Native American mortality disadvantage exceedsmortality (Epidemiology 2023;34: 402–410) Submitted May 18, 2022; accepted January 24, 2023 From the aLa Follette School of Public Affairs, University of Wisconsin— Madison, Madison, WI; bCenter for Demography and Ecology, R University of Wisconsin–Madison, Madison, WI; cMax Planck Institute acial–ethnic disparities in mortality in the United States for Demographic Research, Rostock, Germany; dFederal Institute for Population Research, Wiesbaden, Germany; eSchool of Geography are large and persistent, despite the recent narrowing of and Sustainable Development, University of St Andrews, St Andrews, the gap between Black and White Americans.1 The standard Scotland; and fCenter for Social Data Science and Population Research indicators for measuring and monitoring these inequalities Unit, University of Helsinki, Helsinki, Finland. H.P.A. and M.M. contributed to conceptualization, methodology, writing the include life expectancy and years of life lost. By design, these original manuscript, and review and editing. H.P.A. conducted formal indicators are based either on synthetic populations in which analysis, supervision. E.A. contributed to methodology, formal analysis, age structure is implicitly derived from the mortality schedule data curation, visualization, and review and editing. C.D. contributed to methodology, formal analysis, and review and editing. J.M.H. contributed of the observed population or, in the case of years of life lost, to writing and editing the manuscript. on standard populations. That is, these measures disregard dif- E.A. received funding from the Social Sciences and Humanities Research ferences in the actual population age structures when com- Council (Canada) - Postdoctoral grant No. 756-2019-0768, and the Fonds de recherche du Québec – Société et culture - Postdoctoral grant No. paring mortality experiences across populations. This feature 274299. Support for this research was provided to HPA by the University may be highly beneficial in contexts in which the underlying of Wisconsin-Madison, Office of the Vice Chancellor for Research and differences in real population age structures are considered a Graduate Education with funding from the Wisconsin Alumni Research Foundation. nuisance to the analysis.2 For health policy, however, measures All data and code to fully reproduce the analyses are available at the OSF: that account for population size and structure may be helpful https://osf.io/hj8g3. in providing a signal on where to allocate scarce resources. The authors report no conflicts of interest. Correspondence: Héctor Pifarré i Arolas, La Follette School of Public Affairs, Insofar as mortality disparities inform social and health pol- University of Wisconsin—Madison, Madison, WI 53706. E-mail: hparo- icy priorities, and greater inequalities demand more attention, las@lafollete.wisc.edu. identifying the extent to which some racial groups experience Copyright © 2023 The Author(s). Published by Wolters Kluwer Health, Inc. a mortality disadvantage should aid in guiding policy. This is an open-access article distributed under the terms of the Creative Ignoring the actual age structure of populations can Commons Attribution-Non Commercial-No Derivatives License 4.0 (CCBY-NC-ND), where it is permissible to download and share the work result in misleading conclusions in some contexts. The cur- provided it is properly cited. The work cannot be changed in any way or rent COVID-19 pandemic offers a salient example, illustrated used commercially without permission from the journal. by the following thought experiment. Consider two cruise ISSN: 1044-3983/23/343-402-410 DOI: 10.1097/EDE.0000000000001595 ships with populations from the same country; both cruises 402 | www.epidem.com Epidemiology • Volume 34, Number 3, May 2023 Epidemiology • Volume 34, Number 3, May 2023 US Racial–Ethnic Mortality Gap carry individuals from all ages, but one has a relatively young in the case of life expectancy, synthetic population structures population, while the other is populated mostly by retirees. that are implied by mortality rates. In which ship is a COVID-19 outbreak more threatening? We propose a novel measure for evaluating mortality Given that the risk of complications and death after infection inequalities that accounts for the actual age structures of the increase exponentially with age,3,4 we expect it to be the lat- populations. Our starting premise is that mortality rate differ- ter. However, the loss of life as measured by the life expec- ences are based on mutable social inequities, such as mate- Downloaded from http://journals.lww.com/epidem by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1 tancy reductions would be identical. This is because changes rial deprivation or unequal access to care. Based on this, we in life expectancy are determined by the preoutbreak age-spe- suggest a counterfactual approach that is closely related to cific mortality rates and the age-specific COVID-19–related existing methods, such as indirect standardization and well- AWnYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC4/OAVpDDa8KKGKV0Ymy+78= on 04/25/2023 increase in mortality, which are shared across scenarios. The established decomposition approaches.27 We formulate the actual age structures play no role even though we know that, counterfactual by asking to what extent mortality conditions everything else equal, older populations are bound to be more would improve or worsen for a given racial–ethnic popula- heavily afflicted by the pandemic.5 Thus, an assessment of the tion, given their age composition, if they were to experience loss of life from COVID-19 that is based on life expectancy non-Hispanic White age-specific mortality rates (our base- would not detect this age-structure–related vulnerability. This line). From a technical perspective, we differ from previous shortcoming is known,6 but often not sufficiently acknowl- studies that assess racial mortality inequality using indirectly edged; in our analysis, we demonstrate its implications for the standardized measures by differentially weighting deaths by evaluation of racial–ethnic mortality disparities. remaining life expectancy.28 A key alternative summary measure to life expectancy However, we believe our primary contribution is con- is years of life lost (YLL), which is commonly used to assess ceptual. We explicitly analyze the role of age composition, the relative importance of specific causes of death within a whereas the literature is dominated by approaches that treat population and thus guide public health interventions.7 Years age composition purely as a confounder. Current age-struc- of life lost are the sum of the years between the age at which ture–related adjustments implicitly or explicitly use age death occurs and the age at which we would expect death to structures that differ from those of real populations. This is occur.8,9 Years of life lost measures are directly age standard- clearly the case for measures based on direct standardization, ized for cross-population comparisons, using a shared popula- but it also applies to life expectancy comparisons. The life tion age structure.9 This is not without controversy, as there is table contains an implicit population age structure; that is, the no generally agreed-upon objective way to choose the stan- age structure that would occur for a population with current dard population, and results may vary strongly depending on mortality rates and no growth rate (just replacement). In that the standard,10 affecting not only the magnitude of disparities sense, life expectancy can be interpreted as the mean age at but even their direction.11 The limitations of direct standard- death of that synthetic population. These deviations from real ization have been acknowledged to be particularly relevant in population age structures come at a cost, as illustrated by the racial inequality assessments, given the existing differences examples in this introduction. in age structures across racial–ethnic groups.12,13 Thus, direct We employ a life table approach, similar to much of the standardization is a partial solution at best. literature on racial–ethnic mortality differentials. Accordingly, These two approaches feature prominently in the cur- exposures are defined as the population at risk of dying during rent literature on population-level racial–ethnic mortality an age interval, measured in person years. The main differ- disparities. Life expectancy differences have been frequently ence is that our adjustment procedure then corrects for the used to evaluate US trends in racial–ethnic mortality dispari- age structure of the real population by using its age-specific ties,1,14–16 as well as its geographic patterns.17 This approach exposures and weighting mortality risks accordingly; that is, has also been applied to document the disproportionate our method is exposure-corrected. That is, while other meth- impact of the COVID-19 pandemic on racial–ethnic minori- ods are based on synthetic populations, our correction uses the ties in the United States,18 resulting in greater life expectancy actual population age structure in the definition of the at-risk disparities.19 In turn, years of life lost have been used to assess populations. Exposure adjustments are especially relevant in the cause-specific differentials in mortality that result in the the assessment of racial–ethnic mortality disparities given the overall racial–ethnic mortality disparities. More recently, this substantial differences in age structures across racial–ethnic framework has also been applied to assess loss of life inequali- groups. ties during the pandemic.20,21 This aligns with prior work that We demonstrate this new measure with a case study evaluates contributions of leading causes to racial–ethnic evaluating contemporary mortality inequalities in the United disparities using YLL,22 as well as more targeted work that States among Hispanics, non-Hispanics who are American has focused on specific causes of death and risk factors, such Indian/Alaska Native (AIAN), Asian American, non-Hispanic as alcohol consumption,23 cancer,24 and skin cancer,25 HIV,26 Blacks, and non-Hispanic White. To highlight the magnitude among others. All of the above rely either on direct standard- of the differences among measures, we compare our measure ization, often based on the 2000 US standard population or, of exposure-corrected inequalities to the inequalities estimated © 2023 The Author(s). Published by Wolters Kluwer Health, Inc. www.epidem.com | 403 Pifarré i Arolas et al. Epidemiology • Volume 34, Number 3, May 2023 when using life expectancy and standardized years of life lost. (sYLL), we compute a counterfactual YLL (cYLL), with A’s Our findings suggest that standard demographic indicators age structure and B’s mortality rates. underestimate the mortality inequalities in the United States  cYLLA = mBx · ex · CxA · 100, 000 (4) because they disregard actual population age structures when x comparing mortality experiences across populations. The results based on the novel indicator, which we call the gap Downloaded from http://journals.lww.com/epidem by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1 Then, we measure the mortality gap as the difference adjusted for population structure, suggest that inequalities between cYLL and the actual YLL, that is, the YLL at both A’s are substantially larger than standard demographic methods current mortality rates and age structure of exposures; that is, would imply. AWnYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC4/OAVpDDa8KKGKV0Ymy+78= on 04/25/2023 the GAP (∆cYLL) is the difference between (2) and (4). METHODS ∆cYLLA,B = YLLA − cYLLA = GAPA,B (5) Standard Measurement of the Racial–Ethnic Mortality Gap Thus, a positive GAP (5) shows how many years of life The object of interest of this study, the racial–ethnic are lost (a “disadvantage”) and a negative GAP how many mortality gap, is measured as the difference between given years are gained (an “advantage”) relative to a reference pop- mortality indices across two racial–ethnic groups (A, B). For ulation’s mortality. This age-structure–dependent measure of life expectancy, this is simply: mortality differences can then be used as the foundation to study the racial–ethnic mortality disparity, without arbitrarily ∆eA,B A B 0 = e0 − e0 , (1) choosing a population for (direct) age standardization. G Similar counterfactual approaches are frequently used where e0 stands for the life expectancy at birth of group in demography, epidemiology, and public health. Our pro- G. Another common strategy is based on the years of life lost posal follows the logic of indirect standardization methods.29 (YLL) framework.8,9 While typically YLL are used to assess Indirect standardization measures, such as the standardized cause-related mortality, they can also be utilized to assess the mortality ratio (SMR), compare observed death counts to a mortality gaps. The standard approach calculates YLL per counterfactual based on the actual population age structure death as the difference between the age at death and remaining and the reference population’s mortality rates. We embrace life expectancy at that given age. At a population level, they this approach, which is often but not exclusively used in are often expressed in rates (per 100k):  data-sparse situations, and extend it to consider the age gra- YLLG = mG G x · ex · Cx · 100, 000, (2) dient of the loss of life accrued by deaths. That is, the loss x of life accrued to age-specific deaths in our measure is based on remaining life expectancy, whereas the SMR implicitly where mG G x , ex , Cx are the mortality rates, standard weights each death equally (regardless of age). How can these remaining life expectancy (based on some standard; we use counterfactual mortality rates be understood? Mortality dis- best practices, more details on that in the data section), and parities are rooted in mutable social inequities, such as socio- exposures at age x, respectively. In cross-population compari- economic disparities and racial discrimination. For example, sons a reference standard population (Cxs ) is often used; we the approximately double infant mortality rates Blacks suffer denote this variant of YLL the sYLL. Based on sYLL the mor- compared with Whites30 have been tied to access to care,31 tality gap between two populations is as follows: race-related biases at treatment,32 and other social factors.33 Thus, it is conceivable that disadvantaged racial–ethnic ∆sYLLA,B = sYLLA − sYLLB (3) groups might achieve the lower mortality rates of the more advantaged. This is our counterfactual. A Counterfactual Approach to Mortality Intuitively, there may be a tendency to place more Inequality importance on mortality risks affecting larger fractions of the We propose a counterfactual method to the measure- population. In the infant mortality rate example above, for a ment of mortality disparities. The basic idea, in the spirit of group with a young population with high fertility rates, reduc- indirect standardization methods, consists of measuring the ing infant mortality could represent a greater reduction in disparity between a population A and its reference B as the loss of life than improving mortality rates at the upper end mortality change for A that would result from attaining B’s of the mortality distribution. And the contrary could be true mortality rates while holding constant A’s age structure. We for a relatively older population group, as in the COVID-19 embed this calculation in the standard years of life lost (YLL) example above, where focusing more resources on COVID- approach. We call this indicator the gap adjusted for popula- 19 prevention and treatment could lead to fewer YLL. Thus, tion structure (GAP). We operationalize this idea in the fol- adjusting for the actual age structure of the population may lowing manner. Instead of using direct age-standardized YLL lead to a greater understanding of how to prioritize public 404 | www.epidem.com © 2023 The Author(s). Published by Wolters Kluwer Health, Inc. Epidemiology • Volume 34, Number 3, May 2023 US Racial–Ethnic Mortality Gap health interventions to reduce loss of life. Our measure, the year of age, sex, and bridged race and Hispanic origin from the population structure-adjusted mortality gap, allows us to eval- Underlying Cause of Death data, available through the CDC uate these trade-offs. WONDER Online Databases.35 We defined causes of death following the 10th revision of the International Statistical Comparisons to Other Approaches Classification of Diseases (ICD-10). We used the US standard To compare our results with those from existing meth- population and US population estimates by single year of age Downloaded from http://journals.lww.com/epidem by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1 ods measuring the mortality gap using life expectancy (∆e0) (0, 1, 2,…, 85 years and over), abridged race, Hispanic origin, or standardized YLL (∆sYLL ), we normalize each measure and sex, on 1 July 2019, from the abridged-race intercensal by dividing it by the (race–ethnicity specific) loss of life from estimates of the resident population of the US website.36 We AWnYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC4/OAVpDDa8KKGKV0Ymy+78= on 04/25/2023 leading causes of death. Note that this normalization is not used the WHO’s standard life expectancies by single year of necessary for interpreting the population structure-adjusted age and a 85+ open-ended age interval.8,9 For exposure adjust- mortality gap alone, but this is useful for comparative pur- ment, we used age-specific exposures taken from the Human poses. First, it serves as a reference for the magnitude of the Mortality Database,37 and combined them with the racial– population structure-adjusted mortality gap,2 by comparing ethnic distribution by age to calculate race- and age-specific it with important causes of death with which we are already exposures. This analysis using publicly available data did not familiar. Second, it facilitates the comparison of the gap across require ethical review. metrics that use different units of measure. The importance of these causes of death within a racial–ethnic group is assessed Application using standard approaches; for e0, we use a cause-deleted life We used the population structure-adjusted mortality gap table approach11 and for YLL, we compute the YLL associated to reanalyze contemporary racial–ethnic mortality inequali- with each cause of death.8,9 ties in the United States. We compare non-Hispanic American The population structure-adjusted mortality gap mea- Indian or Alaska Native, non-Hispanic Asian Americans, non- sures mortality gaps through counterfactuals based on exam- Hispanic Blacks, non-Hispanic Whites, and Hispanics (irre- ining actual populations with alternative mortality rates. In spective of race) to non-Hispanic Whites (each group compared contrast, both ∆e0 and ∆sYLL evaluate the mortality dispari- with non-Hispanic Whites). We use circulatory diseases (ICD ties for counterfactuals based on synthetic populations with codes I00-I99), the leading cause of death in the United States, actual mortality rates. In the case of ∆sYLL , it captures the for normalizing disparities. The Table displays all the elements mortality inequalities that would exist should all races/ethnici- of these calculations and the results. We divide the mortality dif- ties share the age structure of a reference population (e.g., the ference calculated according to each approach (Table, column US standard population). In turn, ∆e0-based assessments can 4) by the loss of life from a cause of death (Table, column 5). be given a population interpretation, whereby differences in For Blacks, the normalized mortality disparity is interpreted as life expectancy at birth between specific populations are the follows: the mortality disadvantage of Blacks compared with disparity in mean ages at death in their respective stationary Whites represents X times as much loss of life as circulatory populations. Therefore, both approaches evaluate mortality diseases. That is, from the perspective of Blacks, achieving disparities in synthetic or theoretical populations which may White mortality across all ages would result in an overall reduc- not reflect the disparities that occur in the actual populations. tion in loss of life X times the gain from eliminating circula- Figure 1 illustrates the differences in age structure by tory diseases. Realistically, the former intervention would likely race–ethnicity and gender between the real US population and involve improving in a variety of causes of death. Hispanics the US standard and the life table populations. For instance, and Asian Americans have lower mortality than Whites, so the compared with the actual age distribution of the total White ratio captures the mortality advantage of Hispanics and Asian population in 2019, the proportion of those under one year of Americans. We calculate similar relative metrics based on age is 39% and 27% larger in the US standard and life table ∆e0 and ∆sYLL . In addition to circulatory diseases, we also populations, respectively. Asian American, AIAN, Hispanic, normalize disparities with the following four leading causes and Black actual populations are younger than the standard of death in the United States: cancers (ICD codes C00-D48), and life table populations suggest (more so in the Asian external causes (V01-Y89), respiratory diseases (J00-J98), American and Hispanic cases), whereas the reverse is true for and diseases of the nervous system (G00-G98). Whites. The gender-specific age structures show patterns that are, generally speaking, aligned with those of their respective racial–ethnic groups. RESULTS Figure 2A shows the mortality disparity for all groups Data except Whites (our baseline) based on life expectancy (∆e0 We retrieved life tables for year 2019 by single year of ), the direct age-standardized (using the US standard popu- age, sex, race, and Hispanic origin from the National Center lation) YLL (∆sYLL ), and the population structure-adjusted for Health Statistics (NCHS) Life Tables website’s public mortality gap (∆cYLL). Each point represents the ratio of the files.34 We obtained death counts for 2019 by cause, single racial–ethnic disparity to the gains from removing circulatory © 2023 The Author(s). Published by Wolters Kluwer Health, Inc. www.epidem.com | 405 Pifarré i Arolas et al. Epidemiology • Volume 34, Number 3, May 2023 Asian Black Hispanic Native White 80 60 Female 40 Downloaded from http://journals.lww.com/epidem by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1 20 0 AWnYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC4/OAVpDDa8KKGKV0Ymy+78= on 04/25/2023 80 60 Male Age 40 20 0 80 60 Total 40 20 0 0.2 0.5 1.0 2.0 5.0 0.2 0.5 1.0 2.0 5.0 0.2 0.5 1.0 2.0 5.0 0.2 0.5 1.0 2.0 5.0 0.2 0.5 1.0 2.0 5.0 Age composition relative to the actual population FIGURE 1. Deviations from the Population Actual US Standard Life Table actual age distribution (2019). diseases. We find that, across race–ethnicity and gender, mor- In some cases ∆sYLL and population structure-adjusted tality disparities based on GAP are larger than those captured mortality gap differ substantially. For instance, for Hispanic by ∆e0. men compared with White men, ∆sYLL indicates that eth- The Black/White disparity in ∆e0 is two thirds (0.67) nic disparities are roughly as deadly as circulatory diseases the size of the loss of life expectancy from circulatory diseases (1.01), while the population structure-adjusted mortality gap for the total Black population. However, based on YLL—both indicates that ethnic disparities are considerably larger (1.18). ∆sYLL and population structure − adjusted mortality gap— The extent to which real age structures differ from those of the the racial disadvantage is as deadly as the leading cause of life table or the standard population plays a role in the differ- death (1.07 and 1.15). That is, the Black/White racial disparity ences between methods, with larger deviations in age struc- is 72% (1.15/0.67) larger based on our approach. In the same tures resulting in larger disparities in the resulting gaps. The way, AIAN/White disparities, already deadlier than circula- Asian American, Hispanic, and AIAN populations deviate the tory diseases according to ∆e0 (1.5), increase 65% (2.48/1.5) most (across races/ethnicities) from synthetic age structures, based on the GAP. and we find large differences between our method and existing For racial–ethnic groups with a mortality advantage approaches. over Whites, the disparities are also markedly larger as mea- sured by the population structure-adjusted mortality gap. Gender-specific Results The Hispanic mortality advantage based on ∆sYLL and When interpreting race–ethnicity and gender-specific GAP (1.24 and 1.39) is more than two times larger than what results, we note that the difference-to-cause ratio will depend the ∆e0-based metric implies (0.57). In the case of Asian on both the gender-specific importance of the reference cause Americans, the mortality difference as measured by popula- of death and of the gender racial mortality gap itself. Thus, for tion structure-adjusted mortality gap (3.85) is more than triple a given racial–ethnic group, a given gender may have a larger the amount ∆e0 indicates (1.19), with respect to the loss of life mortality gap, but a smaller difference-to-cause ratio. This is from circulatory diseases. not a limitation intrinsic to the population structure-adjusted 406 | www.epidem.com © 2023 The Author(s). Published by Wolters Kluwer Health, Inc. Epidemiology • Volume 34, Number 3, May 2023 US Racial–Ethnic Mortality Gap TABLE. Life Expectancy (Δe0), Standardized YLL (ΔsYLL), Population Structure-adjusted Mortality Gap (GAP) for: the Racial Gap Between Each Race and Whites (Column 4), the Loss of Life from Circulatory Diseases for Each Race (Column 5), and the Ratio of the Two (Column 6) Mortality Gain from Circulatory Ratio of Mortality Difference to Race Gender Metric Difference Diseases Elimination life Loss Circulatory Diseases Downloaded from http://journals.lww.com/epidem by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1 Asian Female ∆e0 −6.17 5.74 1.07 ∆sYLL −6,452.88 1,624.44 3.97 GAP −6,919.50 1,688.27 4.10 AWnYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC4/OAVpDDa8KKGKV0Ymy+78= on 04/25/2023 Male ∆e0 −7.12 5.53 1.29 ∆sYLL −9,733.46 2,987.21 3.26 GAP −9,690.15 2,718.56 3.56 Total ∆e0 −6.79 5.72 1.19 ∆sYLL −8,073.29 2,232.05 3.62 GAP −8,355.41 2,172.03 3.85 Black Female ∆e0 3.16 6.04 0.52 ∆sYLL 4,875.13 4,841.84 1.01 GAP 5,150.98 5,023.60 1.03 Male ∆e0 5.01 5.66 0.88 ∆sYLL 9,369.15 8,193.86 1.14 GAP 9,425.65 7,294.22 1.29 Total ∆e0 3.99 5.98 0.67 ∆sYLL 6,753.40 6,330.30 1.07 GAP 7,012.41 6,115.05 1.15 Hispanic Female ∆e0 −3.17 5.39 0.59 ∆sYLL −3,350.31 2,170.04 1.54 GAP −2,722.47 1,595.72 1.71 Male ∆e0 −2.73 5.15 0.53 ∆sYLL −3,974.91 3,918.96 1.01 GAP −2,997.18 2,546.89 1.18 Total ∆e0 −3.08 5.42 0.57 ∆sYLL −3,702.54 2,975.08 1.24 GAP −2,875.87 2,069.03 1.39 Native Female ∆e0 6.31 4.40 1.43 ∆sYLL 11,178.17 4,058.96 2.75 GAP 11,214.59 4,087.90 2.74 Male ∆e0 7.69 4.77 1.61 ∆sYLL 16,086.81 7,338.55 2.19 GAP 16,116.01 6,881.70 2.34 Total ∆e0 7.03 4.67 1.50 ∆sYLL 13,494.89 5,599.07 2.41 GAP 13,535.15 5,459.49 2.48 In the case of Asians and Hispanics, the mortality gap is negative (advantage), whereas for Blacks and AIAN it is positive (disadvantage). mortality gap, but rather a consequence of the normalization For this reason, direct comparisons of the mortal- undertaken to facilitate the interpretation of the magnitudes ity difference (Table, column 4) across genders were also with respect to other approaches. This consideration plays a informative. We found, based on the population structure- role in the gender-specific results across races/ethnicities. adjusted mortality gap, that the mortality advantage for © 2023 The Author(s). Published by Wolters Kluwer Health, Inc. www.epidem.com | 407 Pifarré i Arolas et al. Epidemiology • Volume 34, Number 3, May 2023 A Asian Black Hispanic Native 1.07 0.52 0.59 1.43 ∆e0 Female 3.97 1.01 1.54 2.75 ∆sYLL 4.1 1.03 1.71 2.74 GAP 1.29 0.88 0.53 1.61 ∆e0 Downloaded from http://journals.lww.com/epidem by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1 Male 3.26 1.14 1.01 2.19 ∆sYLL 3.56 1.29 1.18 2.34 GAP 1.19 0.67 0.57 1.5 ∆e0 AWnYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC4/OAVpDDa8KKGKV0Ymy+78= on 04/25/2023 Total 3.62 1.07 1.24 2.41 ∆sYLL 3.85 1.15 1.39 2.48 GAP 0.2 0.5 1.0 2.0 5.0 0.2 0.5 1.0 2.0 5.0 0.2 0.5 1.0 2.0 5.0 0.2 0.5 1.0 2.0 5.0 Ratio of the racial gap vs loss of life from circulatory diseases B Circulatory Cancer Female External Respiratory Nervous Circulatory Cancer Male External Respiratory FIGURE 2. Mortality gap based Nervous on life expectancy (∆e0), stan- dardized YLL (∆sYLL ) and the Circulatory Cancer population structure-adjusted mortality gap (GAP). A, The Total External Respiratory racial/ethnic disparity for females, Nervous males, and total in relation to cir- 1 3 10 30 1 3 10 30 1 3 10 30 1 3 10 30 culatory diseases. B, Racial/ethnic Ratio of the racial gap vs loss of life from leading causes disparity for females, males, and total using other causes of death Measure ∆e0 ∆sYLL GAP (top 5 causes) as a reference. Asian Americans and Hispanics was larger for males than disadvantage was also larger based on GAP, 47% (1.29/0.88) for females, and the mortality disadvantage for Blacks and for males and 98% (1.03/0.52) for females. Finally, we found AIAN was more pronounced for males. This is aligned with similar patterns for AIAN, with larger disadvantages for both the findings based on ∆e0 for most racial–ethnic groups, with males, 45% (2.34/1.61), and females, 92% (2.74/1.43). the exception of Hispanics. The Hispanic mortality advantage as measured by GAP (Table, column 4) was larger for males Comparison to Other Leading Causes of Death than for females (−2,997 and −2,722, respectively), whereas In Figure 2B, we normalized by the five leading causes an evaluation based on ∆e0 indicates that males (−2.73) had a of death in the United States to assess whether our results are smaller life expectancy advantage over females (−3.17). idiosyncratic for circulatory disease or demonstrative of a pat- Nonetheless, an analyst may also be concerned with the tern. We find that racial–ethnic disparities based on existing magnitude of the gender-specific racial gap with respect to the approaches (∆e0, ∆sYLL ) are smaller across almost all causes within gender circulatory diseases mortality; that is, the differ- of death. The only exception was external causes, particularly ence-to-cause ratio (Table, column 6). Overall, results based for Hispanics, for which the disparity was larger based on ∆e0 on the population structure-adjusted mortality gap indicate than on our approach. This is because, as with racial–ethnic larger disparities than those based on ∆e0 for both genders. In disparities, loss of life due to external causes is also underesti- that regard, the mortality advantage for Asian Americans over mated by current approaches, especially for Hispanics. This is Whites was 283% larger (4.10/1.07) for females and 176% because it is the combination of a cause with higher mortality larger (3.56/1.29) for males, than the disadvantage as measured at early ages and a considerably younger subpopulation than by ∆e0. Similarly, the Hispanic advantage is 190% (1.71/0.59) the US standard or life table populations suggest (Figure 1). larger for females and 123% (1.18/0.53) larger for males than The size of the mortality difference, as measured by life expectancy-based disparities would imply. For racial–eth- other leading causes of death, varies substantially by race– nic groups experiencing a mortality disadvantage, we also ethnicity and gender. However, the magnitude of the differ- found larger disparities based on GAP. The Black mortality ences we report are similar, across racial–ethnic groups, for 408 | www.epidem.com © 2023 The Author(s). Published by Wolters Kluwer Health, Inc. Epidemiology • Volume 34, Number 3, May 2023 US Racial–Ethnic Mortality Gap the case of respiratory diseases, and even more so for nervous distinct perspectives on racial–ethnic equity in mortality. The diseases (Figure 2). That is, the magnitude of the disparity is underlying notion of equity behind ∆e0-based analysis is that not uniquely high for circulatory diseases. equality will be achieved when any two individuals born in the same birth cohort in the United States, regardless of race–eth- DISCUSSION nicity, have the same life expectancy. Alternatively, we have Using NCHS, CDC, WHO, and HMD data, and a presented a different version whereby equality implies that, Downloaded from http://journals.lww.com/epidem by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1 novel counterfactual-based measure, we studied contempo- given their age structure, no race–ethnicity would be better rary racial–ethnic mortality disparities in the United States. off exchanging their mortality rates with those of any other Our results show that racial–ethnic mortality evaluations that race–ethnicity. The two notions are not equivalent, as we AWnYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC4/OAVpDDa8KKGKV0Ymy+78= on 04/25/2023 account for actual exposures indicate larger mortality dispari- have illustrated in this work, and thus can lead to different ties than analyses based on mortality rates and their implied recommendations for policies that pursue the reduction of dis- age structure, such as life expectancy. Based on our exposure- parities. Given that both approaches have similar data require- corrected measure, the population structure-adjusted mortal- ments and analytical complexity, the preferred approach will ity gap, we find a larger Black/White (72%) and AIAN (65%) depend on the research question. mortality disadvantage, and a greater Asian American (224%) and (144%) Hispanic mortality advantage than what life CONCLUSIONS expectancy-based calculations would imply. These dispari- The purpose of this piece is to introduce a new approach ties are also larger than the results obtained using standard- to measuring mortality disparities. The population structure- ized years of life lost measures, indicating that using real age adjusted mortality gap explicitly incorporates actual age structures instead of standard age structures identifies greater structures in cross-population analyses. We also compare this racial–ethnic mortality disparities in the US context. measure to the disparity found using two common alterna- Our measure complements existing approaches to esti- tives—life expectancy and years of life lost. Other indicators mate mortality trends and disparities; ultimately, the best that quantify additional dimensions of mortality and health approach depends on the question. We have posited that disparities might benefit from insights from this work, such exposures ought to play a central role in understanding popu- as measures of life table-based longevity like the median and lation-level disparities, and as such, the population structure- modal ages at death41 and lifespan inequality indices.42 Other adjusted mortality gap is particularly well suited to studying measures, such as quality adjusted life-years (QALY), often racial–ethnic mortality inequalities. The focus on exposures used in policy evaluations,43 consider the disability status might not always be warranted. Age-specific life expectan- of years lived. Although these measures provide additional cies are commonly used as population-based estimates of insights on racial–ethnic mortality disparities beyond those remaining life years for actuarial calculations, such as the covered by our approach, the exposure-related considerations ones involved in the forecasts of pension expenditures.38 In at the core of our contribution also apply. Exploring the quan- trying to understand and model individual behavior (e.g., titative implications of incorporating exposure corrections savings decisions), individual survival probabilities play a into these measures represents an interesting potential avenue central role.39 Finally, we might want to hold age structures for future research. constant in assessing temporal trends in issues such as the burden of disease, as to assess the improvement of condition- REFERENCES 1. Harper S, MacLehose RF, Kaufman JS. Trends in the black-white life specific mortality rates.40 Our approach is thus tailored for expectancy gap among US states, 1990–2009. Health Aff (Millwood). analyses in which age structures are a fundamental compo- 2014;33:1375–1382. nent, not merely a confounder. 2. Wrigley-Field E. US racial inequality may be as deadly as COVID-19. Proc Natl Acad Sci USA. 2020;117:21854–21856. Similarly to existing approaches, a shortcoming of our 3. Sasson I. Age and COVID-19 mortality: a comparison of Gompertz approach is that we do not consider the effect that changing doubling time across countries and causes of death. Demogr Res. mortality rates could have on age structures. This is clearly the 2021;44:379–396. 4. Goldstein JR, Lee RD. Demographic perspectives on the mortal- case for ∆sYLL , because direct age standardization assumes ity of COVID-19 and other epidemics. Proc Natl Acad Sci USA. that different mortality rates can coexist with an identical 2020;117:22035–22041. age distribution. It might appear that ∆e0 does not suffer 5. Dudel C, Riffe T, Acosta E, van Raalte A, Strozza C, Myrskylä M. Monitoring trends and differences in COVID-19 case-fatality rates using FROM this shortcoming. Indeed, in the calculation of e0, the decomposition methods: contributions of age structure and age-specific age structure is implicitly derived from the mortality rates. fatality. PLoS One. 2020;15:e0238904. However, the implicit age structure of the life table is based 6. Vaupel JW. How change in age-specific mortality affects life expectancy. Population Stud. 1986;40:147–157. exclusively on mortality rates, and thus does not represent a 7. Adam T. Making choices in health: WHO guide to cost-effectiveness anal- realistic approximation of the age structure under alternative ysis. World Health Organization; 2003. mortality distributions. 8. Martinez R, Soliz P, Caixeta R, Ordunez P. Reflection on modern meth- ods: years of life lost due to premature mortality—a versatile and com- Beyond technical considerations, the approaches to prehensive measure for monitoring non-communicable disease mortality. mortality evaluation we have presented also correspond to Int J Epidemiol. 2019;48:1367–1376. © 2023 The Author(s). Published by Wolters Kluwer Health, Inc. www.epidem.com | 409 Pifarré i Arolas et al. Epidemiology • Volume 34, Number 3, May 2023 9. WHO, Geneva. WHO methods and data sources for global burden 25. Ekwueme DU, Guy GP Jr, Li C, Rim SH, Parelkar P, Chen SC. The of disease estimates 2000-2011. Department of Health Statistics and health burden and economic costs of cutaneous melanoma mortality Information Systems. 2013 Nov. by race/ethnicity-United States, 2000 to 2006. J Am Acad Dermatol. 10. Wyper GM, Grant I, Fletcher E, McCartney G, Fischbacher C, Stockton 2011:S133–S143. DL. How do world and European standard populations impact burden of 26. Trepka MJ, Niyonsenga T, Fennie KP, McKelvey K, Lieb S, Maddox disease studies? A case study of disability-adjusted life years (DALYs) in LM. Sex and racial/ethnic differences in premature mortality due to HIV: Scotland. Arch Public Health. 2020;78:1–8. Florida, 2000-2009. Public Health Rep. 2015;130:505–513. Downloaded from http://journals.lww.com/epidem by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1 11. Preston S, Heuveline P, Guillot M. Demography: Measuring and 27. Kitagawa EM. Components of a difference between two rates. J Am Stat Modeling Population Processes. Blackwell Publishers. 2000. Assoc. 1955;50:1168–1194. 12. Krieger N, Williams DR. Changing to the 2000 standard mil- 28. Satcher D, Fryer GE Jr, McCann J, Troutman A, Woolf SH, Rust G. What lion: are declining racial/ethnic and socioeconomic inequalities if we were equal? A comparison of the black-white mortality gap in 1960 in health real progress or statistical illusion? Am J Public Health. and 2000. Health Aff (Millwood). 2005;24:459–464. AWnYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC4/OAVpDDa8KKGKV0Ymy+78= on 04/25/2023 2001;91:1209–1213. 29. Inskip H, Beral V, Fraser P, Haskey J. Methods for age‐adjustment of 13. Thurber KA, Thandrayen J, Maddox R, et al. Reflection on modern meth- rates. Stat Med. 1983;2:455–466. ods: statistical, policy and ethical implications of using age-standardized 30. Driscoll AK, Ely DM. Effects of changes in maternal age distribution and health indicators to quantify inequities. Int J Epidemiol. 2022;51:324–333. maternal age-specific infant mortality rates on infant mortality trends: 14. Harper S, Rushani D, Kaufman JS. Trends in the black-white life expec- United States, 2000-2017. Natl Vital Stat Rep. 2020;69:1–8. tancy gap, 2003-2008. JAMA. 2012;307:2257–2259. 31. Kitsantas P, Gaffney KF. Racial/ethnic disparities in infant mortality. J 15. Schwandt H, Currie J, Bär M, et al. Inequality in mortality between Black Perinat Med. 2010;38:87–94. and White Americans by age, place, and cause and in comparison to 32. Greenwood BN, Hardeman RR, Huang L, Sojourner A. Physician–patient Europe, 1990 to 2018. Proc Natl Acad Sci U S A. 2021;118:e2104684118. racial concordance and disparities in birthing mortality for newborns. 16. Arias E, Johnson NJ, Vera BT. Racial disparities in mortality in the adult Proc Natl Acad Sci USA. 2020;117:21194–21200. Hispanic population. SSM Popul Health. 2020;11:100583. 33. Geronimus AT. The weathering hypothesis and the health of African- 17. GBD US Health Disparities Collaborators. Life expectancy by county, American women and infants: evidence and speculations. Ethn Dis. race, and ethnicity in the USA, 2000-19: a systematic analysis of health 1992;2:207–221. disparities. Lancet. 2022;400:25–38. 34. NCHS. Life tables. Available at: https://www.cdc.gov/nchs/products/ 18. Arias E, Tejada-Vera B, Ahmad F, Kochanek KD. Provisional life expec- life_tables.htm. Accessed 6 October 2022. tancy estimates for 2020. NVSS Vital Statistics Rapid Release: From the 35. NCHS. CDC WONDER online databases portal. Available at: https:// Centers for Disease Control and Prevention, National Center for Health wonder.cdc.gov/. Accessed 6 October 2022. Statistics, National Vital Statistics System. 2021 Jul 1;15: 1–12. 36. NCHS. Abridged-race population estimates. Available at: https://www. 19. Aburto JM, Tilstra AM, Floridi G, Dowd JB. Significant impacts of the cdc.gov/nchs/nvss/bridged_race/216data_documentation.htm. Accessed COVID-19 pandemic on race/ethnic differences in US mortality. Proc 6 October 2022. Natl Acad Sci U S A. 2022;119:e2205813119. 37. HMD. Human Mortality Database. Available at: https://www.mortality. 20. Pierce JB, Harrington K, McCabe ME, et al. Racial/ethnic minority and org/. Accessed 6 October 2022. neighborhood disadvantage leads to disproportionate mortality burden 38. Social Security Administration. The 2021 Annual Report of the Board and years of potential life lost due to COVID-19 in Chicago, Illinois. of Trustees of the Federal Old-Age and Survivors Insurance and Federal Health Place. 2021;68:102540. Disability Insurance Trust Funds. 2021. Available at: https://www.ssa. 21. Xu JJ, Chen JT, Belin TR, Brookmeyer RS, Suchard MA, Ramirez CM. gov/oact/TR/2021/index.html. Accessed 15 November 2021. Racial and ethnic disparities in years of potential life lost attributable to 39. Post T, Hanewald K. Longevity risk, subjective survival expectations, and COVID-19 in the United States: an analysis of 45 states and the District individual saving behavior. J Econ Behav Organ. 2013;86:200–220. of Columbia. Int J Environ Res Public Health. 2021;18:2921. 40. Murray CJ, Vos T, Lozano R, et al. Disability-adjusted life years 22. Aragón TJ, Lichtensztajn DY, Katcher BS, Reiter R, Katz MH. Calculating (DALYs) for 291 diseases and injuries in 21 regions, 1990–2010: a sys- expected years of life lost for assessing local ethnic disparities in causes tematic analysis for the Global Burden of Disease Study 2010. Lancet. of premature death. BMC. 2008;8:116. 2012;380:2197–2223. 23. Shield KD, Gmel G, Kehoe-Chan T, Dawson DA, Grant BF, Rehm 41. Canudas-Romo V. Three measures of longevity: time trends and record J. Mortality and potential years of life lost attributable to alcohol con- values. Demography. 2010;47:299–312. sumption by race and sex in the United States in 2005. PLoS One. 42. van Raalte AA, Sasson I, Martikainen P. The case for monitoring life-span 2013;8:e51923. inequality. Science. 2018;362:1002–1004. 24. Lortet-Tieulent J, Soerjomataram I, Lin CC, Coebergh JWW, Jemal 43. Neumann PJ, Cohen JT, Weinstein MC. Updating cost-effectiveness—the A. U.S. burden of cancer by race and ethnicity according to disability- curious resilience of the $50,000-per-QALY threshold. N Engl J Med. adjusted life years. Am J Prev Med. 2016;51:673–681. 2014;371:796–797. 410 | www.epidem.com © 2023 The Author(s). Published by Wolters Kluwer Health, Inc.