The Burden of Diabetes Mortality by County, Race, and Ethnicity in the U.S., 2000–2019 Free

Diabetes is a major cause of death and morbidity in the U.S. (1). From 1990 to 2021, the age-standardized prevalence of diabetes (type 1 and type 2 combined) increased by 141%, reaching 9,001 per 100,000 in 2021 (2). Although diabetes mortality rates have declined in recent decades—from a high of 27 deaths per 100,000 in 2003 to 22 deaths per 100,000 in 2021 (3)—as a result of improved treatment, approximately 74,000 people died of diabetes in the U.S. in 2021, and a further 310,000 deaths due to other causes (including ischemic heart disease, chronic kidney disease, and stroke, among others) were attributed to high fasting plasma glucose (3,4). Diabetes is also associated with high health expenditures and economic costs. In 2017, the total cost of diabetes in the U.S. was approximately $327 billion (USD), including $237 billion in direct health expenditures and $90 billion because of decreased productivity (5). Type 2 diabetes is far more common than type 1 diabetes; among adults with diagnosed diabetes in 2016, 90.9% had type 2, 5.8% had type 1, and 3.3% had other types of diabetes (6). Studies of total diabetes burden, including the estimates presented in this analysis, therefore primarily reflect type 2–specific outcomes.

Studies have found considerable variation in diabetes mortality across U.S. states and counties (7–9). Variation has also been observed across racial and ethnic populations (1,10,11). In a study of Michigan Medicare beneficiaries with type 2 diabetes, American Indian or Alaska Native (AIAN) and Black individuals had a greater risk of mortality compared with White individuals (33% and 15% higher, respectively, in 2012), whereas mortality risk was similar among Latino and White individuals and lower (by 14%) for Asian individuals (10). Several studies have reported trends in diabetes mortality rates stratified jointly by race and ethnicity and by geographical location (12,13); however, these studies had limited geographical scope (one metropolitan area and the 50 most populous cities, respectively), which may overlook significant variation in diabetes mortality patterns across other regions, especially outside of major cities. This study aims to fill this gap and presents trends in diabetes mortality rates in the U.S. from 2000 to 2019 for five racial and ethnic populations and across counties. By examining county-level data, we aim to identify local disparities that may contribute to differential diabetes outcomes and to provide insights that could support targeted public health strategies and policies. Reporting of such granular data on the temporal and spatial patterning of diabetes mortality by race and ethnicity can help inform and optimize prevention and treatment interventions, ultimately guiding efforts to reduce diabetes deaths in the communities that are most affected.

This analysis used methods for estimating cause-specific mortality by county and racial and ethnic population that have previously been applied to other causes of death and described in detail (14). We briefly summarize these methods and describe their application to diabetes mortality below. This study complies with the Guidelines for Accurate and Transparent Health Estimates Reporting (15) (Supplementary Table 1) and received institutional review board approval from the University of Washington (Seattle, WA). No primary data were collected for this study, and we had no contact with human subjects.

We estimated diabetes mortality rates stratified by county, combined race and Latino ethnicity, and year (2000–2019). We combined counties to create stable geographic units of analysis, which reduced the number of areas analyzed from 3,143 to 3,110. We considered five mutually exclusive racial and ethnic populations: AIAN, Asian or Pacific Islander (Asian), Black, Latino or Hispanic (Latino), and White. Prior to 2011, individuals with less common Asian or Native Hawaiian or Pacific Islander (NHPI) ethnicities were combined in a residual “Other Asian/NHPI” category in the deaths data, which prevented us from considering separate Asian and NHPI populations in this analysis. We refer to the combined Asian and NHPI population as “Asian,” as the estimates for the combined population primarily reflect the experience of the Asian population, which was more than 30 times larger than the NHPI population in 2019 (16).

This study used the cause list and hierarchy from the Global Burden of Diseases, Injuries, and Risk Factors (GBD) 2021 study, a global, systematic effort to measure health loss from more than 300 diseases and injuries (3). While the focus of this paper is on diabetes (including type 1 and type 2), all causes in the GBD cause hierarchy for which there were at least 10,000 deaths in total over the study period were analyzed concurrently to ensure consistency across estimates for different causes of death. Deidentified individual-level death records from the U.S. National Vital Statistics System were tabulated by county, racial and ethnic population (using the “bridged” race imputed by National Center for Health Statistics [NCHS] for individuals with multiple racial identities) (17), age group (0, 1–4 years, 5-year age bands from 5–9 to 80–84 years, and 85+ years), sex, year, and cause of death. Race was missing from 0.62% of death certificates; in these cases, we used the value imputed by NCHS. Latino ethnicity was missing from 0.27% of death certificates; in these cases, we allocated deaths to Latino and non-Latino ethnicity in proportion to the corresponding population sizes by county, race, sex, and year (pooling over time and across counties, as needed, to achieve stable populations). We applied algorithms from the GBD study to reallocate deaths originally assigned an intermediate, immediate, implausible, or insufficiently specific cause as the underlying cause of death (3); this impacted 17.8% of all deaths when considering the third level of the GBD cause hierarchy (the level that includes diabetes). For this analysis of diabetes, we included 1,272,340 deaths where the underlying cause of death was assigned ICD codes E10.1, E10.3–E11.1, E11.3–E12.1, E12.3–E13.1, E13.3–E14.1, E14.3–14.9, P70.2, and R73-73.9 and an additional 147,902 deaths that were reallocated to diabetes via the algorithm described above. Deaths where diabetes was an intermediate or contributing cause, but not the underlying cause, were not included in this definition.

Tabulated deaths data were combined with population estimates by county, race and ethnicity, age, sex, and year from NCHS. We leveraged data on income and population density by county, and data on postsecondary education, poverty, and birthplace in versus outside of the U.S. by county, race, and ethnicity as covariates in the statistical model to improve the predictions of diabetes mortality (but not to adjust for these variables); the data sources and processing methods for these covariates have been previously described (14). Finally, we incorporated existing estimates of racial and ethnic misclassification ratios, defined for each racial and ethnic population as the ratio of the number of deaths where that race and ethnicity were self-reported and the number of deaths where that race and ethnicity were recorded on the death certificate (18).

We used a small-area estimation model to estimate diabetes mortality rates by county, racial and ethnic population, age, and year; details on the model specification, model validation, and model performance have been previously described (14). We then multiplied the modeled mortality rates for each racial and ethnic population by the corresponding misclassification ratios, to adjust the mortality rates for misclassification. Next, we carried out a post hoc calibration on the estimates for each county to ensure consistency in the estimates for different causes of death, and to ensure that adjusting for misclassification did not change the mortality rate for all racial and ethnic populations combined. Finally, we calculated state- and national-level estimates as the population-weighted average of the county-level estimates, and then calculated age-standardized mortality rates using the population distribution from the 2010 Census. To estimate 95% uncertainty intervals (UIs), we simulated 1,000 “draws” from the approximated posterior distribution of the small-area model and carried out all subsequent calculations on each draw; the reported UIs were calculated as the 2.5th and 97.5th percentiles of these draws. We described changes over time as “statistically significant” when the posterior probability that the change was greater than zero was either less than 2.5% or greater than 97.5%. We masked (i.e., did not report) estimates for counties and racial and ethnic populations where the mean annual population size was less than 1,000, as our previous validation analysis found that model performance declined substantially below this threshold (14). Estimates that were not masked (i.e., for counties and racial and ethnic populations with mean annual population greater than 1,000) were referred to as “unmasked.”

Estimates of diabetes mortality by county, racial and ethnic population, year, age, and sex are available for download from the Global Health Data Exchange (https://ghdx.healthdata.org/record/ihme-data/us-diabetes-county-race-ethnicity-2000-2019) and via a user-friendly data visualization: https://vizhub.healthdata.org/subnational/usa. Information about the underlying data sources has been published elsewhere (14). The code used for this analysis is available on GitHub: https://github.com/ihmeuw/USHD.

From 2000 to 2019, the overall age-standardized diabetes mortality rate in the U.S. decreased by 32.1% (95% UI 31.6–32.6), from 28.1 deaths per 100,000 (27.9–28.2) to 19.1 per 100,000 (19.0–19.2) (Fig. 1). Decreases were also observed across all five racial and ethnic populations during this period, with the largest absolute reduction in mortality rates occurring among the AIAN population (decrease of 27.9 deaths per 100,000 [24.0–32.2], from 63.5 [56.7–71.0] in 2000 to 35.6 [32.1–39.4] in 2019), followed by the Black population (decrease of 24.1 [23.3–24.8], from 56.0 [55.3–56.6] to 31.9 [31.5–32.3]), and the Latino population (decrease of 19.8 [18.9–20.6], from 39.5 [38.5–40.4] to 19.7 [19.3–20.2]). Mortality decreased more modestly for the White population (by 6.9 deaths per 100,000 [6.7–7.1], from 24.6 [24.4–24.7] to 17.6 [17.5–17.8] in 2000 and 2019, respectively) and for the Asian population (decreasing by 5.5 [4.8–6.2], from 18.1 [17.4–18.9] to 12.6 [12.1–13.1]). These absolute changes between 2000 and 2019 corresponded with percentage decreases of 44.0% (40.9–46.9), 43.0% (42.1–44.0), 50.1% (48.9–51.3), 28.1% (27.6–28.7), and 30.2% (27.3–33.4) for the AIAN, Black, Latino, White, and Asian populations, respectively. However, this rate of change was not constant throughout the study period: mortality rates for all racial and ethnic populations combined increased 1.9% (1.3–2.5) from 2000 to 2005, decreased 34.3% (33.8–34.7) from 2005 to 2013, and increased 1.4% (0.6–2.3) from 2013 to 2019. Broadly similar patterns were observed for the Asian, Black, Latino, and White populations individually, whereas mortality in the AIAN population continued to decrease after 2013, albeit at a slower rate than during the previous period (2005–2013).

The rank ordering of racial and ethnic populations by diabetes mortality remained unchanged over the study period, but, because the largest absolute mortality rate decreases were in the three populations with the highest mortality rates (AIAN, Black, and Latino), disparities also decreased for each population compared with the Asian and White populations. Differences between highest and lowest national-level mortality rates (among the AIAN and Asian populations, respectively) dropped by 22.4 deaths per 100,000 (18.5–26.8), from 45.4 (38.8–52.6) in 2000 to 22.9 (19.3–26.9) in 2019; this corresponded to a decline in the mortality rate ratio from 3.5 (3.1–3.9) to 2.8 (2.5–3.1). Notably, there was a much larger decrease in the Latino population than the White population from 2000 to 2019, and, as a result, the gap in mortality rates between these two populations was nearly eliminated by the end of the study period: the absolute difference decreased by 12.9 deaths per 100,000 (12.0–13.7), from 14.9 (13.9–15.9) in 2000 to 2.1 (1.6–2.6) in 2019, corresponding to a decline in the mortality rate ratio from 1.6 (1.6–1.6) to 1.1 (1.1–1.1).

Age-standardized diabetes mortality rates varied substantially among counties, but the amount of variation and the spatial patterns differed by racial and ethnic population (Fig. 2). In 2019, within-population variability in mortality rates across counties with unmasked estimates, as measured by the interquartile range (IQR), was highest for the AIAN population (median 31.9 [IQR 18.7–50.3] deaths per 100,000) and lower—in descending order—for the Black (median 33.8 [IQR 27.5–41.3]), Latino (median 16.5 [IQR 12.4–22.6]), White (median 20.6 [IQR 16.6–25.2]), and Asian (median 12.8 [IQR 10.5 to 15.6]) populations. For the AIAN population in 2019, the highest estimated mortality rates (highest 1% of rates [range 107.5–219.3 deaths per 100,000]) were in counties in central Mississippi, southwestern Oklahoma, and eastern South Dakota, and the lowest estimated mortality rates (lowest 1% of rates [range 5.8–6.6]) were in counties in central Texas, northern Alabama, northeastern Illinois, and New York. For the Black population, among counties with the highest mortality rates (highest 1% of rates [range 78.4–124.4]), most (12 of 15) were in northern Louisiana, the Mississippi Delta, and southern Mississippi (with the remainder in northeastern Arkansas, southwestern Georgia, and southwestern Pennsylvania), while those with the lowest mortality rates (lowest 1% of rates [range 11.2–15.8]) were broadly dispersed throughout the upper Midwest states, central California, central Colorado, Hawaii, southern Mississippi, and central Texas. Twelve of 15 counties with the highest rates among the Latino population in 2019 (highest 1% of rates [range 43.1–58.6]) were in western Texas, with the remaining three counties in northeastern New Mexico and southwestern Oklahoma, particularly in the regions of the Texas Panhandle and South Plains, while 6 of 15 counties with the lowest rates (lowest 1% of rates [range 5.2–7.0]) were in Mississippi, with the other nine dispersed throughout Southeastern states, Alaska, and Ohio. Among the White population, the highest mortality rates (highest 1% of rates [range 45.5–69.6]) were mostly spread in the South and Appalachian regions, with notable concentrations in Kentucky (7 of 31 counties), Mississippi (6), Virginia (6), Louisiana (4), and West Virginia (3), whereas 13 of 31 counties with the lowest mortality rates (lowest 1% of rates [range 3.3–8.5]) were in the Rocky Mountain region of Colorado. For the Asian population, both the counties with the highest rates (highest 1% of rates [range 25.4–44.1]) were spread in the Western and Midwestern U.S., and those with the lowest rates (lowest 1% of rates [range 4.5–6.7]) were geographically scattered across Alaska, the Northeast, the Midwest, and the South.

Between 2000 and 2019, 2,826 (91.8%) of 3,079 counties with unmasked estimates (1,788 [58.1%] counties statistically significant) experienced a decline in the diabetes mortality rate for the total population (Fig. 3). Mortality among the White population decreased in a smaller percentage of counties (2,638 [86.5%] of 3,051 counties; 1,511 [49.5%] statistically significant) compared with the AIAN (471 [99.4%] of 474 counties; 358 [75.5%] statistically significant), Asian (662 [99.3%] of 667 counties; 207 [31.0%] statistically significant), Black (1,440 [96.8%] of 1,488 counties; 1,125 [75.6%] statistically significant), and Latino (1,446 [97.8%] of 1,478 counties; 1,075 [72.7%] statistically significant) populations. Furthermore, the magnitude of the relative change in diabetes mortality rates from 2000 to 2019 varied across counties, with the greatest geographical variability observed for the White population (median −22.6% [IQR −33.6% to −9.4%]), followed by the Latino (median −45.0% [IQR −53.0% to −33.5%]), AIAN (median −42.0% [IQR −52.6% to −33.4%]), Asian (median −30.6% [IQR −37.6% to −22.0%]), and Black (median −39.9% [IQR −46.8% to −31.7%]) populations. For the White population, counties where mortality increased between 2000 and 2019 were mainly located in the Southeast, lower Midwest, and Southwest. Notably, 21 of 31 counties with the largest relative increases (highest 1% of increases, range 50.5–124.2%) were clustered in southwestern and southeastern Georgia, and the Delta and central regions of Mississippi. Among the Black and Latino populations, most counties with increasing rates were in the Southeast, especially in southwestern and southeastern Georgia and the Delta and central Mississippi for the Black population (40 of 48 total counties with estimated increases), and in North Carolina and the central and southern parts of Georgia for the Latino population (26 of 32 counties). Counties with the largest percent decreases among the Black population (highest 1% of percent decreases [range 66.5–76.7%]) were also generally concentrated in the Southeast, especially in southeastern Louisiana (11 of 15 counties), whereas the largest declines for the Latino population (highest 1% of declines [range 71.5–81.0%]) were predominantly located in central and southeastern Texas (14 of 15 counties).

As was the case nationally, rates of change in diabetes mortality rates at the county level differed across years in the study period (Fig. 4 and Supplementary Table 2). Approximately half of counties with unmasked estimates had decreasing rates from 2000 to 2005 among the Asian, Black, and Latino populations, while only around a quarter of counties had decreases among the AIAN and White populations (most with relatively modest decreases). In contrast, from 2005 to 2013, declines were observed across nearly all counties with unmasked estimates for all five racial and ethnic populations, and the magnitude of the decline was generally more substantial. For the Black and White populations, the relatively few counties with increasing mortality rates during this period (2005–2013) were located mainly in Georgia and Mississippi; these same states also contained many of the counties with the largest observed increases from 2000 to 2005. From 2013 to 2019, mortality declined in just over half of counties for the Black population; in a minority of counties for the Asian, Latino, and White populations; and in a large majority of counties for the AIAN population. While mortality increased in a relatively large proportion of counties both from 2000 to 2005 and from 2013 to 2019 for the Asian, Black, Latino, and White populations, the counties that experienced relatively large increases were often different between these two periods.

Our analysis of diabetes mortality in the U.S. from 2000 to 2019 by county, race, and ethnicity shows substantial declines in mortality rates for all five racial and ethnic populations included in the study. Encouragingly, the populations with the highest mortality rates—AIAN, Black, and Latino—also experienced the largest declines, nearly eliminating the disparity between the Latino and White populations, and substantially reducing the disparity for the AIAN and Black populations compared with the Asian and White populations. However, both geographical and racial and ethnic disparities persisted, and progress on reducing mortality slowed after approximately 2013. These findings highlight both notable public health victories and ongoing challenges. The detailed estimates presented here, characterizing spatial and temporal variation in diabetes mortality rates and associated health inequities, can provide an evidence base through which to thoroughly explore the drivers of these trends—information that is crucial to designing and implementing more effective, focused interventions moving forward.

Improved management of type 1 and type 2 diabetes is likely an important factor in the declines in diabetes mortality rates observed across all racial and ethnic populations and most counties from 2005 to 2013 (19). Modifications to diabetes diagnostic criteria in the 1990s lowered the diagnostic threshold for fasting blood glucose, alongside the introduction of glycosylated hemoglobin (HbA_1c) into clinical use, thus leading to more individuals being diagnosed with type 2 diabetes and initiating treatment at earlier stages (20). Furthermore, the introduction of the drug metformin in the U.S. in 1994 was revolutionary in controlling blood glucose and decreasing glycosylated hemoglobin (HbA_1c) levels (21) in patients with diabetes and has since been shown to reduce the incidence of type 2 diabetes by up to 31% among people with prediabetes (22). Alongside metformin use has been the emergence of different types of insulin treatments (i.e., short-acting versus long-acting) to effectively control blood glucose throughout the day, as well as the more recent market introduction of novel therapies like glucagon-like peptide 1 (GLP-1) agonists for blood glucose stabilization and comorbidity (specifically, obesity) mitigation (23). Such changes in practice and advancements in treatment and diagnosis, although applied at the individual level, have likely contributed to the declines observed in most counties and nationally because of their widespread application and integration as standard of care for people with diabetes. Furthermore, as type 2 diabetes substantially increases the risk of cardiovascular disease, it is likely that more widespread use of statins for cardiovascular disease prevention and management has also been reflected in decreasing diabetes mortality rates (24). Finally, there has been further benefit to diabetes outcomes from the increasingly aggressive management of hyperglycemia, high blood pressure, and high cholesterol levels (all of which are risks for complications among individuals living with diabetes) in recent decades (25). Despite decades-long progress, however, control of blood glucose, lipid, and blood pressure levels has regressed among patients with diabetes since 2010, subsequently increasing diabetes-associated complications (26). This more recent shift in disease management may partially underlie the distinct changes in diabetes mortality rate trends observed both nationally and at the county level after 2013, highlighting the need for novel clinical and public health approaches in mitigating diabetes burden moving forward. Importantly, efforts to mitigate diabetes burden should not focus solely on clinical interventions but should also address upstream, often community-wide factors such as social determinants of health that contribute to disparities in both diabetes incidence and poor outcomes including death.

In addition to the slowing of progress toward the end of the study period, it is critical to address the racial and ethnic disparities and geographical disparities that remained in 2019. Variation in diabetes burden across racial and ethnic populations has previously been linked to a combination of factors, including differences in the prevalence of diabetes; prevalence of underlying risk factors (in the case of type 2 diabetes) and comorbidities such as obesity, mental health conditions, commercial tobacco smoking, limited physical activity, and poor diet; and inequities in access to and quality of health care (27). These same factors vary geographically and likely contribute to geographical disparities in diabetes mortality, although more investigation is needed to determine the specific factors driving diabetes mortality in any particular location. Obesity, in particular, has been a key factor affecting trends in diabetes morbidity and mortality, as it acts as a risk for both diabetes onset and diabetes-associated complications (because of the increased risk for comorbidities like heart and liver disease) and has been precipitously rising across the country in recent decades (28). An analysis from GBD 2021 identified high BMI as the biggest contributor to an estimated 321% increase in prevalence of diabetes in the U.S. from 1990 to 2021 (2). Obesity burden in the U.S. also varies by race and ethnicity: in 2017 and 2018, the age-standardized prevalence of BMI ≥30 was 49.6% among Black adults, 44.8% among Latino adults, 42.2% among White adults, and 17.4% among Asian adults (29). Obesity is a complex disease that includes numerous environmental, socioeconomic, and genetic factors, but many cases are at least partially influenced by low levels of exercise and poor diet (i.e., nutrient-deficient, calorie-rich, and with high levels of saturated fat, salt, and sugar), and thus these facets are common targets for modification within obesity treatment plans (30). There is evidence that the risk of developing and dying from type 2 diabetes may be lowered through interventions focusing on exercise and diet, with data from the Diabetes Prevention Program and Outcomes Study showing that improvements to these factors may reduce the incidence of type 2 diabetes by 58% over 3.2 years among individuals with prediabetes (31). At present, there is a great need for widespread education and supportive resources to reduce the burdens of obesity and diabetes, especially among vulnerable populations.

Differences across racial and ethnic populations in access to health resources, including those supporting diabetes care, are likely driven by numerous longstanding systemic, socioeconomic, and psychosocial factors. The AIAN population, for example, continues to bear the legacy and ongoing harms of colonization and structural racism, contending with lower socioeconomic status and more limited access to health care than other racial and ethnic populations (27). As evidenced by the extremely high rates of diabetes mortality among AIAN individuals shown in this study, there is a critical need to address the inequities experienced by this population moving forward, including allocating resources that promote healthy diet and investing in systems to provide reliable, culturally sensitive, and geographically accessible medical care. Black and Latino populations similarly suffer from a lack of access to adequate health care related to structural inequities engendered by racism, including lower average socioeconomic status and higher uninsured rates than the White population (13,32,33). It is notable that the Latino population experienced higher diabetes mortality than the White population, given that this is the opposite of the trend observed for most other causes of death, where the Latino population has substantially lower mortality (14). It is not entirely clear why diabetes has a different pattern than most other causes of death, although it is possible that factors such as lower socioeconomic status and lower access to and quality of health care are especially pertinent for diabetes, a disease where effective management is possible but can be resource- and time-intensive to achieve. One study of adults with health insurance found White patients generally had higher access to diabetes medication and, subsequently, better disease management than other racial and ethnic populations, suggesting barriers beyond lack of health insurance (34). Additionally, residential segregation concentrates other racialized communities in areas with fewer health-promoting resources—areas that, in many cases, are also food deserts (35). The consequences of racism have additionally been apparent within the medical system, with less money and focus directed toward health care in marginalized communities, and a major underrepresentation of community members among health professionals (36). Explicit and implicit bias have generated a lack of trust between marginalized populations and the medical system (36), with serious negative consequences on health outcomes, diabetes awareness, and glycemic control.

There are several limitations to our work, some of which are detailed in previous related research based on the same methodological framework used here (14). First, the deaths, population, and covariates data that underlie the present analysis are subject to error, which may, in turn, lead to inaccuracies in the estimated mortality rates. Second, the mortality rate estimates are uncertain, as quantified by the 95% UIs. This uncertainty is typically greater for smaller populations and for racial and ethnic populations for whom the adjustment required for misclassification of race and ethnicity on death certificates is larger and less reliable. Third, the broad racial and ethnic categories we used to stratify the data inevitably hide important within-population heterogeneity. For example, previous studies have detailed considerable differences in diabetes incidence and prevalence within Asian and Latino populations, with lower prevalence observed for Korean and Japanese populations compared with other Asian populations and with Pacific Islander populations, and with lower incidence and prevalence for Central and South American, Cuban, and Dominican populations compared with Mexican and Puerto Rican populations (11,37,38). Finally, some deaths due to diabetes may be coded as other causes such as cardiovascular disease, thus underestimating the mortality rates due to diabetes (39).

Substantial improvements in diabetes outcomes and disparities across racial and ethnic populations have been achieved over the past two decades. However, diabetes mortality rates remained disproportionately high among certain racial and ethnic populations and geographical locations in 2019, and mortality rate decreases across all races and ethnicities slowed in recent years throughout much of the country. A better understanding of the myriad contributors to the earlier improvements as well as more recent stalled progress and unequal burden can provide valuable insights into the design of effective strategies for clinical and public health practice, policy, and research moving forward. As the U.S. population ages and medical costs increase, special focus on diabetes prevention is needed to ensure equitable health outcomes, regardless of racial and ethnic identity or location.

Funding. This study was funded by the Intramural Research Program, National Institute on Minority Health and Health Disparities, U.S. National Institutes of Health (contract #75N94023C00004), and the Intramural Research Program, National Institute on Minority Health and Health Disparities; National Heart, Lung, and Blood Institute; Intramural Research Program, National Cancer Institute; National Institute on Aging; National Institute of Arthritis and Musculoskeletal and Skin Diseases; Office of Disease Prevention; and Office of Behavioral and Social Sciences Research, U.S. National Institutes of Health (contract #75N94019C00016).

The views expressed are those of the authors and should not be construed to represent those of the U.S. National Institutes of Health or the federal government.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. L.D.-L. and A.H.M. were responsible for the study concept and design. Y.O.K. and M.M.B. extracted and processed the data inputs. P.K., C.A.S., and D.O.S. wrote the computer code and designed and carried out the statistical analyses with input from L.D.-L., H.N., Z.L., P.K., Y.O.K., and M.M.B. prepared the tables and figures. H.N. wrote the first draft of the paper with assistance from Z.L., K.C., and S.A.M. All authors contributed to writing subsequent versions of the manuscript, critically reviewed the methods and results, and approved the final version of the manuscript. W.L.M.-K., F.D., A.H.M., and L.D.-L. managed the project. All authors provided intellectual input into aspects of this study. L.D.-L. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Handling Editors. The journal editors responsible for overseeing the review of the manuscript were Elizabeth Selvin and Alka M. Kanaya.