Racialized Economic Segregation and Breast Cancer Mortality among Women in Maryland

Breast cancer is the second leading cause of cancer-related deaths among Black/African American (AA) women (1). While breast cancer mortality rates are decreasing, Black-White racial disparities in breast cancer mortality rates are increasing (1). Data suggest that racial mortality disparities are due to the higher prevalence of aggressive breast cancer subtypes, late stage of presentation (2, 3), and the high prevalence of comorbidities among Black/AA women (4–6). These mortality disparities are also due to racially patterned differences in access to resources in part due to race-based segregation (7–11), which is compounded by persistent socioeconomic inequalities by race in the United States.

Residential racial and economic segregation remain upstream factors that influence access to equitable healthcare, independent of individual-level risk factors (12), including for breast cancer care (13, 14). For example, unequal access to opportunities and resources such as employment, wealth, income, housing, and education can significantly influence access to high-quality healthcare services, such as cancer prevention/screening, early detection, and treatment services (1, 15). Additionally, access to goods and services and limitations of the built environment (e.g., green spaces, inaccessible or nonexistent sidewalks) can directly affect cancer health outcomes and exposure to cancer risk factors across the cancer continuum (16). While several studies have examined the contribution of residential racial segregation to the observed disparities in breast cancer mortality (7–11, 17), these studies did not explicitly consider how racial composition intersected with the economic conditions of these communities.

The Index of Concentration at the Extremes (ICE) is a segregation metric that reflects the extent to which an area's population is concentrated into extremes of deprivation and privilege. The ICE was initially introduced in the field of social sciences as a measure of economic residential segregation and was expanded to racial and racialized economic segregation, to capture the extremes of high-income White communities and low-income Black communities (18–20). Therefore, ICE captures both racial and economic elements of residential segregation at the same time.

Utilizing data sources from the Maryland Cancer Registry (MCR) and statewide census-level data, we examined the relationships between residential segregation (racial and economic), Black/White race, prognostic/initial treatment factors, and risk of breast cancer mortality among patients with breast cancer. Maryland is an ideal setting to explore these relationships, given its large Black/AA population (30% of residents) that is nearly double the national average (14%; ref. 21). Of the approximately 4,500 women each year that are diagnosed with invasive breast cancer in the state of Maryland, one-third identify as Black/AA (22). The population in Maryland also has varying levels of economic status, as Prince George's County, Maryland has one of the highest median household incomes in the United States according to 2019 data ($84,920 compared with Baltimore City's $50,379; ref. 23). We hypothesized that patients with breast cancer diagnosed in Maryland living in less versus more privileged census tracts across social dimensions of race and income would have higher risk of breast cancer mortality.

The MCR collects data for all cancer diagnoses among residents of the state, including patient demographics, clinical and diagnostic information, and treatment. The MCR routinely links to additional data sources to capture missed cancers and obtain additional information on previously reported cases. Each year the MCR identifies potentially unreported cancer cases by matching the MCR master database with Maryland death certificates that indicate that a patient had cancer. Facilities and physicians are then followed up for information. The registry also links to the Maryland Vital Statistics, Social Security Death Index (SSDI), and the National Death Index (NDI) for information on vital statistics including cause of death. For the present study, a total of 59,230 women diagnosed with breast cancer in Maryland during 2007 to 2017 were identified from the MCR. The study was approved by the Institutional Review Board (IRB) at the Maryland Department of Health and was deemed exempt by the IRB at the Johns Hopkins Bloomberg School of Public Health.

To be included in the study analysis women age ≥18 years had to be diagnosed with primary breast cancer as their first cancer between 2007 and 2017 (excluded n = 10,525). The analysis was then restricted to non-Hispanic Black (NHB) and non-Hispanic White (NHW) women for the study's research question (excluded n = 3,788). Women diagnosed with in situ breast cancer (n = 7,532), or missing age at diagnosis (n = 2), or follow-up time (n = 1,295) were excluded. In addition, 1,022 women were excluded due to missing data associated with variables needed to create the socioeconomic index (i.e., zip code, census tract). The final analytic population included 35,066 women (24,540 NHW; 10,526 NHB).

The primary exposures were the census tract level ICE measure which was estimated using equations from Scally and colleagues (18–20, 24) and data from the U.S. Census American Community Survey 2011 to 2015, which had the greatest overlap with the year of diagnosis among this study population. The description of these equations for the ICE measures are found in Supplementary Table S1. Economic segregation was defined as the difference in the number of persons living in households earning ≤ $25,000 compared with the number of persons living in households earning ≥ $100,000 as a proportion of the total population with household income information. Racial segregation was defined as the difference in the number of Black persons compared with the number of White persons as a proportion of the total persons with race/ethnicity information. Racialized economic segregation was defined as difference in the number of Black persons in households earning ≤ $25,000 compared with White persons in households earning ≥ $100,000 as a proportion of the total persons with income and race/ethnicity information. The segregation indices were calculated in reverse of previous use and range from −1 to 1, such that a value of −1 indicates concentrated privilege (i.e., advantage) and a value of 1 indicates concentrated deprivation (i.e., disadvantage). Quintiles for segregation were constructed based on total number of cases to create equally sized groups and the quintile cutoffs for the segregation indices are presented in Supplementary Table S2. For consistency, quintile 5 is interpreted as the least privileged whereas quintile 1 is interpreted as the most privileged group (reference).

The MCR collects self-reported race from the demographics or diagnostic parts of a patient's record. Race is entered by the facilities at time of diagnosis, and the MCR uses motor vehicle records to enter race when this data is missing. Race was categorized as NHB/AA and NHW (referent), hereto referred to as Black and White, as a representation of differing sociopolitical experiences. The following variables were also collected from the demographic and diagnostic parts of the patients’ records and were evaluated for this current analysis: age at diagnosis, type of initial cancer treatment (surgery only; surgery + radiation; surgery + chemotherapy; surgery + chemotherapy and radiation; no surgery; and treatment unknown), and tumor characteristics including estrogen receptor (ER)/progesterone receptor (PR) status, HER2 status, breast cancer Surveillance Epidemiology and End Results (SEER) Summary stage (localized, regional, distant, unknown), and tumor grade (I, II, III–IV, unknown). HER2 receptor status was also available, although only for 71% of cases, so triple-negative breast cancer was also evaluated [ER-negative (ER⁻), PR-negative (PR⁻), and HER2-negative (HER2⁻)].

Vital statistics information was ascertained by the MCR through linkage to the Maryland Vital Statistics, SSDI and NDI. Data for date and cause of death were most complete through November 5, 2019 from Vital Statistics. The primary cause of death was coded using the International Classification of Diseases-10 (ICD-10) codes for breast cancer–related deaths (C50).

Demographic, clinical, and tumor characteristics of the study population were compared by race as counts and proportions. The census tract level segregation indices were summarized by county (Supplementary Table S3) and mapped by census tract using percentile ranking in Maryland. Percentile ranking maps the segregation indices by their ranking in comparison with other census tracts as opposed to the absolute value of the segregation index. Survival time was calculated as time since diagnosis to mortality or date of last linkage to the death registries. HRs and 95% confidence intervals (CI) were estimated using Cox proportional hazards regression, accounting for clustering at the census tract level with robust standard error specification. We evaluated the proportional hazards assumption by visualizing the cumulative hazards over time and testing an interaction with time in the Cox model. While the cumulative hazards generally appeared proportional over time, there was evidence of nonproportional hazards using an interaction with time. Thus, the HRs are interpreted as average effect over time. As a sensitivity analysis, we also reported the overall HRs by ≤ 5 years and > 5 years.

Age was considered a potential confounder and was adjusted for using natural cubic splines with knots at the fifth, 35th, 65th, and 95th percentiles. Subgroups were also evaluated by overall median age at diagnosis (<60 years, ≥60 years), adjusting for age continuously within each analysis. Tumor stage, hormone receptor status, tumor grade, and treatment were considered potential mediators and not confounders and were therefore not adjusted for in models; however, these variables are also prognostic subgroups of interest, so supplemental analyses were additionally reported stratified by these variables. To evaluate the effects of hormone/endocrine therapy, we restricted data to women with ER-positive (ER⁺) breast cancer. HER2 receptor status was missing for 29% of women, so stratified analyses were not conducted by this subtype. To test for interactions across subgroups, models with and without interactions terms between the strata variable and segregation index were compared using the likelihood ratio test. Since stratified models essentially allow each variable within the model to have a different relationship with the outcome in each stratum, the models used to test the interactions also included interaction terms between age and the strata variable for age-adjusted models as well as between race and the strata variable for age- and race-adjusted models. Because the likelihood ratio test does not allow for clustered variance, the likelihood ratio tests were run on models without accounting for clustering as the standard errors were similar with and without clustering. All analyses were conducted using Stata Version 14.1. All tests were two-sided and P values < 0.05 were considered statistically significant.

The data analyzed in this study are available from the MCR. Restrictions apply to the availability of these data, which were used under license for this study. Data are available from the authors upon reasonable request with the permission of MCR.

Among the 35,066 women in the study, there were over 216,295 person-years of follow-up and 3,739 deaths due to breast cancer. Table 1 describes demographics, clinical and diagnostic factors, and tumor characteristics of the women diagnosed with breast cancer, overall and stratified by race. The median age at breast cancer diagnosis overall was 61 years [interquartile range (IQR) = 52–71]. Black women were younger at diagnosis (median = 58 years, IQR = 49–67) compared with White women (mean = 61 years, IQR = 52–71). The median follow-up time since diagnosis was 5.9 years (IQR = 3.3–8.8).

Tumor characteristics and breast cancer treatments also differed by race. Black women were diagnosed with later stage and higher grade breast cancers compared with White women. Black women had higher prevalence of ER⁻/PR⁻ tumors compared with White women (26.4% vs. 14.6%). Among women with ER⁺ breast cancer, Black women were less likely to receive hormone therapy (51.2%) than White women (60.5%). HER2 status combined with ER and PR status was only available for 69% of women, among which 21.4% of Black women had triple-negative breast cancer (ER⁻/PR⁻/HER2⁻) in comparison with 10% of White women. Black women were also more likely to receive the combination of surgery, chemotherapy, and radiation (23%) compared with White women (15.8%) and less likely (18.7%) to receive the combination of surgery and radiation compared with White women (29.9%).

The census tract segregation indices are summarized with boxplots by Maryland County in Fig. 1. The overall median census-tract economic, racial, and racialized economic segregation across Maryland was −0.20, −0.47, and −0.18, respectively, all leaning towards greater privilege (below 0). Baltimore County was made up of the largest number of census tracts (N = 211) while Garrett County and Somerset County had the least number of census tracts (N = 7). Howard County (−0.54) had the highest economic privilege, Garrett County (−0.97) had the highest racial privilege, and Calvert County (−0.41) had the highest racialized economic privilege. Baltimore city had the lowest economic (0.23), racial (0.62), and racialized economic (0.24) privilege. Figure 2 describes each index visually by census-tract across Maryland, with greater economic segregation (decreasing privilege) centered around census-tracts near Baltimore city, northwest Maryland, southeast Maryland, and smaller pockets around the Washington DC area. Greater racial segregation was observed near the Baltimore city and Washington DC areas. Lastly, racialized economic segregation was greatest again near the Baltimore city and Washington DC areas but also in northwest and southeast Maryland.

The age-adjusted HR and 95% CI for the association between each segregation index and breast cancer mortality, overall and by race and age are shown in Table 2. Additional supplemental analyses stratified by prognostic factors (Supplementary Fig. S1; Supplementary Tables S4 and S5) and time since (≤5 years, >5 years; Supplementary Table S6) were similar across strata. For each segregation index, the HRs are in comparison with the most privileged (quintile 1).

Overall, women in the least privileged quintile of economic segregation had 1.67 times the hazard (95% CI, 1.50–1.86) of breast cancer mortality than women in the most privileged quintile. The associations for breast cancer mortality comparing the least with the most economically privileged were similar by race with a HR of 1.43 (95% CI, 1.19–1.72) and 1.45 (95% CI, 1.26–1.66) for Black and White women, respectively.

The increased hazard of breast cancer mortality associated with living in the least economically privileged areas compared with the most economically privileged areas was stronger for women aged less than 60 years (HR, 2.01; 95% CI, 1.71–2.36) versus women aged ≥ 60 years (HR, 1.44; 95% CI, 1.25–1.66; P_interaction = 0.04; Table 2). In supplemental analyses, models were adjusted for race and results were slightly attenuated although still statistically significant, but the interaction by age was no longer significant (Supplementary Table S4).

Overall, the hazard of breast cancer mortality was 1.57 times as high (95% CI, 1.42–1.74) for women in the least privileged quintile of racial segregation compared with the most privileged quintile (Table 2). The association was similar by race, although among Black women the association was not statistically significant and with wide confidence intervals, while the association was only statistically significant for quintile 4 among White women (HR, 1.16; 95% CI, 1.03–1.31).

The association between racial segregation and breast cancer mortality differed by age and stage. The HR comparing the least with the most racially privileged was 1.83 (95% CI, 1.56–2.14) among women aged less than 60 years and 1.41 (95% CI, 1.23–1.62) among women aged ≥ 60 years (P_interaction = 0.03; Table 2). In supplemental results presented for prognostic factors adjusted for age, the association between racial segregation and breast cancer mortality only differed by stage (P_interaction = 0.02; Supplementary Table S4). In supplemental models additionally adjusted for race, the results were strongly attenuated, and interactions were no longer significant (Supplementary Table S5). Only quintile 4 (the second to least privileged) was still significantly associated with increased breast cancer mortality overall and across age.

Overall, the hazard of breast cancer mortality was 1.84 times as high (95% CI, 1.64–2.06) for the least privileged quintile of racialized economic segregation compared with the most privileged quintile (Table 2). The associations for racialized economic segregation were similar and increased across quintiles among Black women, but not statistically significant and with wide confidence intervals. The association differed by race, such that the least privileged quintile of racialized economic segregation compared with the most privileged quintile was associated with 1.20 (95% CI, 0.90–1.60) times the hazard of breast cancer mortality for Black women as opposed to 1.66 (95% CI, 1.41–1.95) times as high for White women (P_interaction = 0.04).

The association for racialized economic segregation and breast cancer mortality also differed by age (Table 2). Comparing the least racially and economically privileged with the most racially and economic privileged, the association was stronger for women aged less than 60 years (HR, 2.17; 95% CI, 1.83–2.57) versus women aged ≥ 60 years (HR, 1.62; 95% CI, 1.40–1.88; P_interaction = 0.004). In supplemental analyses additionally adjusting for race, results were attenuated; however, the associations still suggested increased breast cancer mortality for decreasing racialized economic privilege (Supplementary Table S5).

Our study examined associations between racialized economic segregation and breast cancer mortality in a cohort of Black and White women diagnosed with invasive breast cancer in Maryland. This is the first study, to our knowledge, to evaluate the associations between ICE metrics and breast cancer mortality by subgroups of race and age. Overall, women in living in neighborhoods with higher concentrations of lower incomes and/or with more Black residents had nearly twice as high the hazard of breast cancer mortality in comparison with women living in neighborhoods with higher concentrations of higher incomes and/or more White residents. In associations stratified by race, only the association between economic segregation and breast cancer mortality was statistically significant among Black women. Age was a significant effect modifier for associations between economic, racial, and racialized economic segregation and breast cancer mortality, with stronger HRs ranging between 1.83 to 2.17 among younger women (<60 years at diagnosis).

While other studies have examined the impact of residential segregation on breast cancer mortality using different metrics [i.e., isolation index (7), measures of socioeconomic status and urbanization (25), location quotient of racial residential segregation (9), racial index of dissimilarity (8), Information Theory Index (10), and “hyper-segregation” (11)], we chose to use the ICE. To date, ICE metrics have only been utilized in a few breast cancer studies: one study comparing ER status in breast cancer cases reported with SEER (20), another study measuring breast cancer survival in Milwaukee, Wisconsin (26), and a study of cancer incidence, including breast cancer in Massachusetts (27). The single ICE metric which combines race/ethnicity and income as a measure of racialized economic segregation is a validated measure of place-based health disparities (18–20, 24, 28).

ICE encapsulates an aspect of social inequity that would be missed if we only considered inequities by the percentage of the population among certain incomes or race/ethnicity and highlights unequal group relationships (24). These otherwise hidden inequities in Maryland can be graphically depicted, as illustrated in Fig. 2. Unsurprisingly, areas of Maryland where racialized economic segregation were greatest (Fig. 2) have the highest age-adjusted female breast cancer mortality rates as described by the 2019 Maryland Department of Health Cancer Data report (2012–2016): Baltimore City (27 per 100,000); Prince George's County near Washington DC (25.1 per 100,000); and Worcester County in southeast Maryland (30.7 per 100,000; ref. 29).

We stratified results by race to explore racial differences in the associations between racialized economic segregation and breast cancer mortality. Black communities generally experience the highest degree of residential segregation of any U.S. group (30, 31). Blacks are more likely to live in segregated areas for numerous reasons, including factors associated with racial discrimination, immigrant settling patterns, and economic disparity (11, 30, 32). In turn, residential segregation tends to limit social, economic, and educational opportunities and resources and is linked to increased poverty in these minority communities (33). Health outcomes can also be negatively impacted by segregation through exposure to substandard housing, lack of access to medical services, and social isolation (12, 30, 34). The only statistically significant association observed among Black women was for the association between economic segregation and breast cancer mortality and results were similar among White women. We did find a significant interaction with individual race and racialized economic segregation with increasing breast cancer mortality among White women living in less privileged areas and no association among Black women. We found no interaction between individual race and racial segregation. Russell and colleagues also explored interactions with individual race and did not find race to be a significant modifier in the association between residential racial composition and breast cancer mortality among women diagnosed with breast cancer in Georgia (22). In additional models, we adjusted for race and found that race attenuated point estimates, most strongly for racial segregation. These results suggest individual race and residential racial composition may be interrelated in the Maryland context.

We observed significant effect modification by age, with a steeper ICE gradient among women under age 60 as compared with women ages 60 and older. This finding may suggest that access to healthcare and insurance status, specifically Medicare that is available to adults as of age ≥ 65 years, reduces disparities in breast cancer mortality arising from racialized economic segregation. In line with this hypothesis, in a study among seniors in the SEER–Medicare database (66–85 years of age), Haas and colleagues did not find a significant association between Black segregation and mortality (HR = 1.03; 95% CI, 0.87–1.21). However, this study was also unable to account for residential socioeconomic status (7). To our knowledge, our findings among younger women have not been explored in other studies of breast cancer mortality and residential segregation and they indicate that age should be considered as a significant factor in this association.

This study had several strengths and limitations. We examined the impact of racialized economic segregation on breast cancer mortality among a large cohort of patients with invasive breast cancer of which 30% were Black. We had sufficient power to evaluate associations among subgroups by race, prognostic factors (i.e., age, hormone receptor status, stage), and treatment modalities. We used ICE metrics that more accurately represent the extremes of economic and racialized segregation in America than other metrics of social inequality. We utilized census tracts for these ICE metrics which might capture the social experience more accurately than county-level metrics. We did not adjust for covariates as we hypothesized that these factors would function as mediators in our analyses, and therefore confounding beyond age could be present if these factors were not actual mediators. We also lacked behavioral risk factor data for risk factors such as physical activity and obesity which could be impacted by racial and economic segregation (35, 36). Evaluation of these factors could be important for further identification of at-risk groups.

In conclusion, our results suggest that breast cancer survival disparities exist in Maryland among women residing in the least privileged census tracts with lower income households and higher proportions of Black residents. We observed significantly stronger associations between racialized economic segregation and breast cancer mortality among younger women. Our findings for ICE were more extreme than those of racial segregation. We believe that these studies that only looked at segregation by race may have underestimated the effects of segregation on cancer outcomes and the intersection between race and socioeconomics should be considered as a critical driver of cancer health disparities. Our study findings provide new insights into the racial disparities in breast cancer mortality observed among women diagnosed with invasive breast cancer in Maryland.

A.E. Connor reports grants from American Cancer Society (ACS) during the conduct of the study; other support from ACS outside the submitted work. M. Kaur reports grants from National Institute of Diabetes and Digestive and Kidney Diseases (training grant T32DK062707) during the conduct of the study; in addition, M. Kaur's work for this study was conducted while M. Kaur was affiliated with Johns Hopkins University. However, at the time of submission, M. Kaur was employed by Flatiron Health Inc., an independent subsidiary of Roche, and received stock ownership in Roche as part of the employment. Flatiron Health and Roche were not involved in this study. L.T. Dean reports personal consulting fees from Weight Watchers, International; however, this work is wholly unrelated to the work published here. No disclosures were reported by the other authors.

The manuscript contents are solely the responsibility of the authors and do not necessarily represent the official views of the Centers for Disease Control and Prevention or the Department of Health and Human Services.

A.E. Connor: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. M. Kaur: Conceptualization, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. K.E. Dibble: Writing–original draft, writing–review and editing. K. Visvanathan: Writing–original draft, writing–review and editing. L.T. Dean: Writing–original draft, writing–review and editing. J.H. Hayes: Resources, data curation, writing–original draft, writing–review and editing.

This work was supported by the ACS (MRSG-19-010-01-CPHPS awarded to A.E. Connor). We would also like to acknowledge support by the NCI Cancer Center Support Grant (P30 CA006973). Cancer data was provided by the MCR, Center for Cancer Prevention and Control, Department of Health, with funding from the State of Maryland and the Maryland Cigarette Restitution Fund. The collection and availability of cancer registry data is also supported by the Cooperative Agreement NU58DP006333, funded by the Centers for Disease Control and Prevention.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.