Racial segregation and disparities in breast cancer care and mortality

Authors

  • Jennifer S. Haas MD, MSPH,

    Corresponding author
    1. Division of General Medicine and Primary Care, Department of Medicine, Brigham and Women's Hospital, Boston, Massachusetts
    2. Department of Medicine, Harvard Medical School, Boston, Massachusetts
    • Division of General Medicine and Primary Care, Brigham and Women's Hospital, 1620 Tremont Street, Boston, MA 02120–1613
    Search for more papers by this author
    • Fax: (617) 732–7072

  • Craig C. Earle MD,

    1. Department of Medical Oncology, Dana–Farber Cancer Institute, Boston, Massachusetts
    Search for more papers by this author
  • John E. Orav PhD,

    1. Division of General Medicine and Primary Care, Department of Medicine, Brigham and Women's Hospital, Boston, Massachusetts
    2. Department of Medicine, Harvard Medical School, Boston, Massachusetts
    Search for more papers by this author
  • Phyllis Brawarsky MPH,

    1. Division of General Medicine and Primary Care, Department of Medicine, Brigham and Women's Hospital, Boston, Massachusetts
    Search for more papers by this author
  • Marie Keohane,

    1. Division of General Medicine and Primary Care, Department of Medicine, Brigham and Women's Hospital, Boston, Massachusetts
    Search for more papers by this author
  • Bridget A. Neville MPH,

    1. Department of Medical Oncology, Dana–Farber Cancer Institute, Boston, Massachusetts
    Search for more papers by this author
  • David R. Williams PhD, MPH

    1. Department of Society, Human Development, and Health, Harvard School of Public Health, Boston, Massachusetts
    Search for more papers by this author

Abstract

BACKGROUND.

Questions have existed as to whether residential segregation is a mediator of racial/ethnic disparities in breast cancer care and breast cancer mortality, or has a differential effect by race/ethnicity.

METHODS.

Data from the Surveillance, Epidemiology, and End Results–Medicare database on white, black, and Hispanic women aged 66 to 85 years with breast cancer were examined for the receipt of adequate breast cancer care.

RESULTS.

Blacks were less likely than whites to receive adequate breast cancer care (odds ratio [OR], 0.78; 95% confidence interval [CI], 0.71-0.86). Individuals, both black and white, who lived in areas with greater black segregation were less likely to receive adequate breast cancer care (OR, 0.73; 95% CI, 0.64-0.82). Black segregation was a mediator of the black/white disparity in breast cancer care, explaining 8.9% of the difference. After adjustment, adequate care for Hispanics did not significantly differ from whites, but individuals, both Hispanic and white, who lived in areas with greater Hispanic segregation were less likely to receive adequate breast cancer care (OR, 0.73; 95% CI, 0.61-0.89). Although Blacks experienced greater breast cancer mortality than whites, black segregation did not substantially mediate the black-white disparity in survival, and was not significantly associated with mortality (hazards ratio, 1.03; 95% CI, 0.87-1.21). Breast cancer mortality did not differ between Hispanics and whites.

CONCLUSIONS.

Among seniors, segregation mediates some of the black-white disparity in breast cancer care, but not mortality. Individuals who live in more segregated areas are less likely to receive adequate breast cancer care. Cancer 2008. © 2008 American Cancer Society.

Several studies have documented that black and Hispanic women are less likely than white women to receive adequate treatment for breast cancer,1–4 and experience greater mortality.5, 6 Racial differences in treatment are not fully explained by differences in clinical and nonclinical factors such as stage, age, comorbidity, health insurance, region, and socioeconomic status.1, 4, 7–11 Residential segregation, the physical separation of 1 racial/ethnic group from others, may adversely affect access to high-quality medical care.12, 13 Residential segregation is more pronounced for blacks than any other racial/ethnic group in the US.12 Segregation may be an important cause of racial/ethnic disparities in care and outcomes.12, 14, 15 For instance, neighborhood racial and ethnic composition may account for some of the disparities in access to medical care.13 We sought to further elucidate this relationship between residential segregation and the adequacy of care and mortality for women with breast cancer. To our knowledge, the role of segregation on disparities in cancer treatment has not been explored. The purpose of this study was to examine whether racial segregation mediates disparities in breast cancer treatment and mortality among US seniors, or has a differential effect for different racial/ethnic groups.

MATERIALS AND METHODS

Data

This analysis uses data from the Surveillance, Epidemiology, and End Results (SEER)-Medicare file.16 The population covered by SEER is comparable to the US population with regard to poverty and education, but tends to be somewhat more urban and has a higher proportion of foreign-born persons than the general US population. These data include information regarding cancer stage, primary tumor site, and patient demographics, and are linked to Medicare claims data by the National Cancer Institute. Data were available for individuals diagnosed with cancer from 1992 to 2002. A restricted access version of these data was obtained so that the characteristics of each individual's census tract of residence, from the 1990 US Census, could be appended. This study was reviewed and approved by the Institutional Review Board of Brigham and Women's Hospital.

Study Sample

Women aged 66 to 85 years, whose race/ethnicity was reported as white, black, or Hispanic, and who were diagnosed with American Joint Committee on Cancer (AJCC) stage I, II, or IIIA breast cancer as a single primary cancer from 1992 to December 2002 were included in this analysis (N = 70,541). As is typical with analyses based on SEER-Medicare data, we excluded women younger than 66 years, so that we would have at least 1 year of Medicare claims to assess comorbidity.17 We also removed women who had health maintenance organization (HMO) coverage and those who were not covered by both Parts A and B for at least 1 year before and 1 year after diagnosis with breast cancer, to ensure that Medicare billing claims were complete (n = 22,508). We excluded women older than 85 years because of more clinical uncertainty regarding the management of these women. Because our analysis was focused on the role of residential segregation, we excluded an additional 167 individuals with a missing census tract identifier. The final sample included 47,866 women.

Variables

Our outcome measure was whether women with breast cancer received adequate breast cancer care. We focused on a previously developed measure of the adequacy of breast cancer care using SEER-Medicare data.1 Adequate care was defined as receipt of effective surgery (either mastectomy or breast conserving surgery [BCS] with radiation therapy) within 6 months of diagnosis, and an assessment of estrogen receptor (ER) status at the time of surgery. BCS with radiation therapy and mastectomy have equivalent survival.18 BCS without radiation is considered to be inadequate therapy, as this treatment is associated with a higher risk of recurrence. Women who received both BCS and mastectomy (both within 6 months of diagnosis) were classified as having a mastectomy (n = 10,679). Assessment of ER status is necessary to determine appropriate chemotherapy. We did not examine the use of chemotherapy, as this was not included as an established indicator of prior breast cancer care.1

To measure residential segregation, we used the isolation index, a measure of the probability that a member of 1 racial/ethnic group will be in contact with members of this same group (as opposed to whites), compared with residential neighbors in the same census tract calculated using the population within 1990 census tracts and counties.19, 20 It is calculated separately for blacks and Hispanics. For instance, the isolation index for blacks within a county = Σ(bi/btotal) × (bi/Ti); in which i = 1 of N census tracts in the county, bi = the number of blacks in the census tract i, btotal = the total number of blacks in the county, and Ti = the total population of census tract i.21 We divided large counties with populations greater than approximately 500,000, such as Los Angeles, into smaller, municipal areas based on locally defined neighborhoods, and calculated the isolation indices for these municipalities based on the census tract populations within these regions (unpublished data). We took this approach because of a growing social science literature that suggests that these areas play a meaningful role in the structure of social processes associated with place of residence, including segregation.22 Municipal areas were defined using information from local urban planning departments and census maps. There are 463 counties in the dataset; 36 of these were separated into 697 municipal areas.

The isolation index ranges from 0 to 1.0, with a higher number indicating greater segregation.23 For example, a county with a black isolation index of 0.6 suggests that a black individual in that county would be very isolated, having a 60% chance of having only other blacks as neighbors. The median household income of the census tract was also included as an independent variable (categorized as <$20,000, $20,000-39,999, $40,000-$59,999, and ≥$60,000).

Individual-level independent variables included age at diagnosis (continuous), race/ethnicity (white, black, Hispanic), marital status (married, not married), cancer stage (I, II, IIIA, based on the modified AJCC), tumor size (continuous in centimeters), Charlson comorbidity index (categorized as 0, 1, 2, ≥3),24 whether an individual was of “low income” (based on eligibility for state assistance with Medicare premiums and copayments), region of residence (Northeast, Midwest, South, West) and year of diagnosis. Missing tumor size (n = 2262) was imputed by the average tumor size for each race/stage.

Analysis

Characteristics of black and Hispanic women were compared with characteristics of white women using chi-square tests. To examine the effect of race/ethnicity and segregation on receipt of adequate care we used multilevel logistic regression models, clustered by county, using SAS statistical software (version 9; SAS Institute, Cary, NC). Initial models were adjusted for age at diagnosis, race/ethnicity, marital status, comorbidity, eligibility for state buy-in coverage, cancer stage, tumor size, year of diagnosis, census region, and the median household income of the census tract of residence. If a significant racial/ethnic effect was found, then the segregation index was entered into the model, and if the index decreased the racial/ethnic effect, then segregation was considered to be a mediator. We examined separate models for blacks compared with whites, and Hispanics compared with whites. To examine the effect of race/ethnicity and segregation on breast cancer mortality, we used Cox proportional hazard models, clustered by county, with SUDAAN version 9 (RTI International, Research Triangle Park, NC) and the same variables described above. Survival was defined as the interval from the date of diagnosis to the Medicare date of death from breast cancer, and if that was missing, SEER date of death. Patients alive at the end of the study period (December 31, 2003) or who died from other causes were censored at that point; the time interval from their date of diagnosis to the end of the study was included in the survival analysis.

RESULTS

We compared the demographic characteristics of individuals included in our sample to those who were excluded because of HMO coverage or limited claims data. Individuals included in the sample were slightly more likely than those excluded to be disadvantaged (rate of ever having qualified for state buy-in coverage of 15.6% vs 15.0%). Individuals included were somewhat less likely to be black than those who were excluded (6.0% vs 7.0%), but were much less likely to be Hispanic (3.6% vs 8.7%).

Compared with white women, black and Hispanic women were younger, less likely to be married, and more likely to have ever been eligible for state buy-in coverage (Table 1). Black and Hispanic women had more comorbidity than whites. Black and Hispanic women were more likely to be diagnosed at a later stage and with a larger tumor than white women. Overall, only 62.6% of women received adequate breast cancer care (mastectomy or BCS with radiation, and assessment of ER status), and black and Hispanic women were less likely than white women to have received this measure of adequate breast cancer care. Specifically, minority women were less likely to have received BCS with radiation, or assessment of ER status. There was no difference in the use of mastectomy between the 3 racial/ethnic groups. Blacks were less likely than whites to survive 5 years after their diagnosis, whereas there were no differences in survival between whites and Hispanics. Whites lived in areas with the highest median household income, and blacks the lowest. Blacks lived in areas with greater black isolation than whites or Hispanics, and Hispanics lived in areas of greater Hispanic isolation than whites or blacks. Blacks were more likely than whites and Hispanics to live in the South or the Midwest. Hispanics were more likely than whites and blacks to live in the West. Black isolation was greatest in the South (mean isolation index 0.36) and lowest in the West (mean isolation index 0.11). Hispanic isolation was greatest in the West (mean isolation index 0.23) and lowest in the Midwest (mean isolation index 0.02).

Table 1. Characteristics of Sample by Race/Ethnicity
 BlackHispanicWhite
  • ER indicates estrogen receptor.

  • *

    P value was derived using the chi–square test; difference between black and white: ≤.0001.

  • P value was derived from the chi–square test; difference between Hispanic and white: <.0001.

  • Data were missing for marital status (130 black individuals, 52 Hispanic individuals, and 1455 white individuals), tumor size (189 black individuals, 57 Hispanic individuals, and 2000 white individuals), and median household income (10 black individuals, 13 Hispanic individuals, and 238 white individuals).

  • §

    P value was derived from the chi–square test; difference between Hispanic and white: <.001.

  • P value was derived from log–rank test comparing black and white (<.0001).

  • P value was derived from the Wilcoxon test; difference between black and white, Hispanic and white: <.0001.

  • #

    P value was derived from the Wilcoxon test; difference between black and white: <.0001.

  • **

    P value was derived from the Wilcoxon test; difference between Hispanic and white: <.0001.

No.2875170443,287
Individual Characteristics   
Age at diagnosis, y*   
 66–7024.1%25.6%20.9%
 71–7529.1%33.8%29.3%
 76–7927.0%23.5%27.5%
 80–8519.8%17.1%22.2%
Married*26.2%39.6%45.9%
Comorbidity*   
 057.6%63.9%71.9%
 124.9%23.1%19.3%
 28.7%6.9%4.9%
 ≥38.8%6.1%4.0%
Ever eligible for state buy–in coverage*43.6%47.7%12.5%
Stage of disease at diagnosis*   
 I45.8%52.6%59.5%
 II49.3%44.3%37.8%
 IIIA4.9%3.2%2.7%
Tumor size, cm*   
 ≤252.1%61.1%67.1%
 2–4.9933.1%30.6%24.7%
 >58.2%4.9%3.5%
Received adequate breast cancer surgery and assessment of ER status*52.0%58.4%63.4%
Mastectomy44.9%45.5%44.5%
Breast–conserving surgery with radiation*§22.8%28.1%32.1%
Assessment of ER status*75.2%78.1%82.3%
5–y mortality from breast cancer17.1%9.5%8.3%
Area Characteristics   
Median household income/census tract of residence$30,594$40,198$47,932
Median black isolation index (interquartile range)#0.60 (0.40–0.88)0.05 (0.02–0.16)0.06 (0.01–0.20)
Median Hispanic isolation index (interquartile range)**0.05 (0.01–0.26)0.41 (0.22–0.58)0.04 (0.02–0.18)
Census Region*   
Northeast13.5%11.7%18.8%
South26.1%1.9%11.2%
Midwest34.8%2.5%25.7%
West25.6%83.9%44.3%

After adjustment for individual characteristics and median household income of the census tract, but not segregation, blacks were less likely than whites to receive adequate care (odds ratio [OR], 0.72; 95% confidence interval [CI], 0.65-0.78). Black segregation mediated some of the disparity in receiving adequate care. As black segregation increased, women, both black and white, were less likely to receive adequate breast cancer care (Table 2) (OR, 0.73; 95% CI, 0.64-0.82). Segregation reduced the black-white disparity in adequate care by 8.9% (unadjusted OR of 0.72 for segregation vs 0.78 adjusted). In a model that adjusted only for demographic and clinical characteristics (age, marital status, comorbidity, tumor size, cancer stage, prior Medicaid coverage, and year of diagnosis), 9.9% of the black-white disparity in the adequacy of breast cancer care was explained by these variables. The model that adjusted for segregation and median household income of the census tract in addition to the demographic and clinical characteristics explained 25.3% of the black-white disparity in receiving adequate breast cancer care.

Table 2. Multivariate Analysis: Individual Characteristics Associated With Adequate Care*
 Adequate Care
Black/WhiteHispanic/White
  • *

    Data are shown as the odds ratio (95% confidence interval).

Race/ethnicity  
WhiteReferenceReference
 Black0.78 (0.71-0.86) 
 Hispanic 0.99 (0.88-1.10)
Age at diagnosis, y0.98 (0.98-0.99)0.98 (0.98-0.99)
Stage at diagnosis  
 Stage IReferenceReference
 Stage II1.12 (1.07-1.18)1.12 (1.07-1.18)
 Stage IIIA1.44 (1.25-1.67)1.55 (1.33-1.80)
Comorbidity  
 0ReferenceReference
 10.99 (0.95-1.05)0.99 (0.95-1.04)
 20.81 (0.75-0.89)0.83 (0.76-.91)
 ≥30.87 (0.78-0.95)0.85 (0.77-0.93)
Marital status  
 MarriedReferenceReference
 Not married0.99 (0.95-1.04)0.99 (0.95-1.04)
Ever eligible for state buy-in  
 NoReferenceReference
 Yes0.89 (0.84-0.95)0.88 (0.83-0.93)
Segregation index  
 Black0.73 (0.64-0.82) 
 Hispanic 0.73 (0.61-0.89)

Although Hispanics were less likely than whites to receive adequate care after adjustment for demographic characteristics, comorbidity, cancer stage, and tumor size (OR, 0.82; 95% CI, 0.74-0.91), this disparity did not persist in the multilevel models that accounted for residential segregation (OR, 0.99; 95% CI, 0.88-1.10) (Table 2). In a model that adjusted only for demographic and clinical characteristics (age, marital status, comorbidity, tumor size, cancer stage, prior Medicaid coverage, year of diagnosis), 1.9% of the Hispanic-white disparity in the adequacy of breast cancer care was explained by these variables. The model that adjusted for segregation and median household income of the census tract in addition to the demographic and clinical characteristics explained 9.1% of the Hispanic-white disparity in receiving adequate breast cancer care. Greater Hispanic segregation was associated with a lower likelihood of receiving adequate care, for both Hispanic and white women (OR, 0.74; 95% CI, 0.61-0.89). Older women, women with more comorbidity, those at early stage, and those who were ever eligible for state buy-in insurance were also less likely to receive adequate care.

Blacks experienced greater breast cancer mortality than whites, even after adjusting for whether they received adequate care (adjusted hazards ratio [HR], 1.44; 95% CI, 1.27-1.63). Adequate care was associated with lower mortality (HR, 0.82; 95% CI, 0.76-0.87). Black segregation did not substantially mediate the black-white disparity in survival, and was not significantly associated with mortality (HR, 1.03; 95% CI, 0.87-1.21). Breast cancer mortality did not differ between Hispanics and whites (HR, 0.94; 95% CI, 0.78-1.13).

DISCUSSION

This results of the current study confirm that many women do not receive adequate care for breast cancer and that there are black-white disparities in breast cancer care and mortality.1–5, 11 Women who live in more segregated areas were less likely to received adequate breast cancer care. This work also suggests that residential segregation may mediate some of the black-white disparity in breast cancer care, but not mortality. We did not find Hispanic-white disparities in breast cancer care or mortality in the fully adjusted models, although disparities in care were apparent in models that only controlled for individual-level characteristics.

Several studies suggest that breast cancer treatments are underutilized by black women relative to white women, after accounting for differences in cancer stage.2, 3, 25–28 Racial differences in the use of breast cancer treatment may be related to differences in access to specialized providers,29, 30 racial bias in referral patterns or treatment recommendations,31, 32 preferences for treatment,33, 34 communication or understanding of risks and benefits,35 and trust in the health care system or research.36, 37 Segregation has been demonstrated to have adverse economic, political, and social consequences.38 We hypothesized that segregation may limit access to specialized cancer providers or to information or understanding of treatment options. Segregation may influence preferences for care, or beliefs regarding the efficacy and side effects of cancer treatment.33 This work suggests that segregation for blacks may adversely affect access to breast cancer care. To the best of our knowledge, prior work has not examined the role of segregation on the use of cancer treatment.

Segregation may be associated with a decline in the adequacy of breast cancer care for whites and blacks because of more limited access to specialized providers in more segregated areas. Our work is consistent with the observation that women living in high-poverty Zip Codes are less likely to receive radiation following BCS than are women living in Zip Codes with lower levels of poverty.4 Our work is also consistent with a recent study of patients with end stage renal disease showing that time to transplantation was longer among both black and white patients in more segregated areas, and that dialysis facilities located in these areas may be of lower quality.39 Prior work suggests that black and white patients are treated by different physicians, and that physicians who care for black patients may have fewer clinical resources and are less likely to be board certified.40 As blacks are much more likely to live in segregated areas than whites, our findings may be explained by the location of these physicians in more segregated areas. In contrast to the decline in the use of adequate breast cancer care with increasing segregation, segregation was not a mediator of disparities in breast cancer mortality among these seniors between blacks and whites. It is not surprising that the effects of segregation were more limited for Hispanics than for blacks, as blacks are more segregated than other racial/ethnic groups in the US.41

Our work has several limitations. Although we were able to adjust for the socioeconomic status of the area in which an individual lived, we had limited information concerning individual socioeconomic status (ie, eligibility for state buy-in coverage for Medicare). There is a growing portfolio of effective, increasingly personalized breast cancer treatments beyond the ones measured here. We focused on a previously developed measure of breast cancer care that would be applicable to the vast majority of women with breast cancer. SEER-Medicare data have been found to be reasonably accurate for assessing the use of surgery, radiation therapy, ascertainment of ER status, and surveillance mammography.42–45 As we examined the care of seniors, our findings may not be generalizable to younger women. Disparities in care may be larger for younger women, particularly since this senior population was all covered by Medicare.46 As has been done in other analyses using these data, we excluded individuals with HMO coverage and those with less than 1 year of claims before or after diagnosis so that comorbidity and treatment could be adequately assessed.17 Again, this may limit the generalizability of our findings, although the racial/ethnic and socioeconomic characteristics of those included versus excluded were similar except for Hispanics. Our analyses are observational and should not be considered causal. Ascertainment of Hispanic ethnicity is likely not complete in SEER or Medicare files, meaning that some Hispanics were most likely misclassified.

Among seniors, segregation may mediate some of the black-white disparity in breast cancer care but not mortality. We did not find disparities in breast cancer care between Hispanics and whites. Segregation was adversely associated with the adequacy of care for all women. Future research should examine mechanisms for these associations, like differences in the availability of specialized providers or the dissemination of health information, so that appropriate interventions can be designed.

Acknowledgements

We thank Joan Warren, PhD for her helpful comments on an earlier version of this article. We thank Philip J. Ethington, PhD for sharing methods for creating measures of residential segregation in municipal areas.

Ancillary