Racial Disparities in the Use of Radiotherapy After Breast-Conserving Surgery: A National Medicare Study

Abstract

BACKGROUND:

In prior studies, the use of standard breast cancer treatments has varied by race, but previous analyses were not nationally representative. Therefore, in a comprehensive, national cohort of Medicare patients, racial disparities in the use of radiotherapy (RT) after breast-conserving surgery (BCS) for invasive breast cancer were quantified.

METHODS:

A national Medicare database was used to identify all beneficiaries (age >65 years) treated with BCS for incident invasive breast cancer in 2003. Claims codes identified RT use, and Medicare demographic data indicated race. Logistic regression modeled RT use in white, black, and other-race patients, adjusted for demographic, clinical, and socioeconomic covariates.

RESULTS:

Of 34,080 women, 91% were white, 6% were black, and 3% were another race. The mean age of the patients was 76 ± 7 years. Approximately 74% of whites, 65% of blacks, and 66% of other-race patients received RT (P < .001). After covariate adjustment, whites were found to be significantly more likely to receive RT than blacks (odds ratio, 1.48; 95% confidence interval, 1.34–1.63 [P < .001]). Disparities between white and black patients varied by geographic region, with blacks in areas of the northeastern and southern United States demonstrating the lowest rates of RT use (57% in these regions). In patients age <70 years, racial disparities persisted. Specifically, 83% of whites, 73% of blacks, and 78% of other races in this younger group received RT (P < .001).

CONCLUSIONS:

In this comprehensive national sample of older breast cancer patients, substantial racial disparities were identified in RT use after BCS across much of the United States. Efforts to improve breast cancer care require overcoming these disparities, which exist on a national scale.

Postoperative radiotherapy (RT) is generally considered a standard component of therapy for patients diagnosed with early invasive breast cancer who undergo breast-conserving surgery (BCS). Cumulative results from randomized trials have demonstrated local control and survival benefits with the addition of RT after BCS.^1–4 However, despite the evidence supporting the use of RT after BCS, a recent analysis of treatment patterns in breast cancer patients enrolled in the North American Fareston versus Tamoxifen Adjuvant (NAFTA) trial suggested that racial disparities may exist in RT use after BCS. Specifically, this study found that postoperative RT was administered in 82% of white patients but only 70% of black patients, although the difference was not found to be statistically significant.⁵

Other studies of breast cancer patients have also suggested racial disparities in the receipt of standard treatments for locally advanced and regional disease, including differences among white and nonwhite patients in the use of RT, surgery, and chemotherapy.^5–15 Results from many of these studies suggest that nonwhite patients may more frequently encounter barriers to receiving standard care. However, prior analyses were conducted on limited samples, including singlestate, single-center, or clinical trial patient cohorts, and studies of population-based cohorts have not to our knowledge been nationally comprehensive. Consequently, the true magnitude of and regional variation in racial disparities in breast cancer treatment across the United States remain uncertain.

Quantifying the magnitude of racial disparities in the use of standard treatments across the nation is an important step toward identifying and ultimately reducing barriers to high-quality breast cancer care. Therefore, using data from a comprehensive, national sample of Medicare patients, we sought to quantify racial differences in the use of RT after BCS for women with invasive breast cancer. In addition, we sought to determine whether there was geographic variation across the United States in the association between race and RT use.

MATERIALS AND METHODS

National Medicare Data Set and Study Sample Derivation

The national Medicare data set includes comprehensive claims data for all Medicare beneficiaries in the United States. Files contain data collected by Medicare for reimbursement of healthcare services for each beneficiary and include institutional (inpatient and outpatient) and noninstitutional (physicians or other providers) final action claims.¹⁶ The University of Texas M. D. Anderson Cancer Center’s institutional review board approved use of the Medicare database for this study.

We used the following algorithm to identify patients with incident breast cancer, which we based on a prior validated algorithm using claims data.^17,18 Our initial study population was comprised of 853,273 women who had any diagnosis of invasive breast cancer in 2003, defined as an International Classification of Diseases-Ninth Revision (ICD-9) diagnosis code of 174. Because this denominator would have included incident and prevalent cases in 2003, as well as potentially disease-free patients undergoing breast cancer screening, we applied the following algorithm to refine our study sample to identify incident cases only. From our initial study population, we included 83,611 patients aged ≥65 years who underwent BCS between January 1, 2003, and December 31, 2003. From this sample, we excluded 17,872 patients who did not have at least 2 claims (on different dates) specifying a diagnosis of invasive breast cancer between January 1, 2003, and December 31, 2004 (at least 1 claim must have occurred during 2003), with no more than 6 months between the date of BCS and the earliest breast cancer diagnosis claim date. To exclude the prevalent cases, we excluded 10,362 patients who had a breast cancer-related diagnosis or procedure claim between January 1, 2002, and December 31, 2002. To ensure that BCS was intended to be the primary cancer-directed surgery, we excluded 13,470 patients who underwent mastectomy within 3 months of BCS. To limit our sample to patients with early stage invasive breast cancer, we then excluded 1082 patients who had ≥2 claims specifying metastatic breast cancer from 3 months before to 3 months after the diagnosis date. To improve sample homogeneity, we also excluded 2964 patients who were receiving Medicare coverage because of end-stage renal disease or disability. Finally, to ensure we had complete claims information to determine patients’ cancer treatment course and comorbidities, we excluded 3781 patients who lacked Medicare Part A or B coverage or who had intermittent health maintenance organization coverage in the 9 months after or in the 1 year before their breast cancer diagnosis date (of these patients, 2143 had incomplete information in the year before diagnosis because they were aged <66 years). For our study, the breast cancer diagnosis date was considered the date of the earliest claim for a diagnosis of breast cancer. This left a final sample size of 34,080 patients. The complete list of claims codes applied in our algorithm can be found in Table 1.

Table 1.

Claims Codes

Variable	ICD-9 Code	CPT Code
Breast-conserving surgery	85.20, 85.21, 85.22, 85.23	19120, 19125, 19126, 19160, 19162
Breast cancer-related diagnosis and procedure claims
Biopsy	85.1–85.19	19000, 19001, 19100, 19101, 19110, 19112
Breast-conserving surgery	As above	As above
Mastectomy	85.33–85.48	19180–19255
Lymph node dissection	40.3	38525, 38740, 38745
Radiotherapy	92.2–92.29, 0330*, 0333*	77400–77499, 77520–77525, 77750–77799
History of breast cancer	V10.3	—

ICD-9 indicates International Classification of Diseases, Ninth Revision; CPT, Current Procedural Terminology.

^*Revenue Center code.

RT and Race

Patients were considered to have received breast RT if a claim for RT (Appendix A) occurred within 9 months of the breast cancer diagnosis date. This definition has been validated in prior studies of breast cancer patients.^19–24 In addition, for patients classified as receiving RT, completion of the therapeutic course was defined as at least 3 complete weeks of therapy (5 treatments per week). Patient race was determined using theMedicare denominator file, which contains demographic information on each beneficiary. In this file, race and ethnicity data are based on the patients’ self-report. For our analysis, we categorized patients as white, black, and other (nonwhite, nonblack) race based on a prior study of the sensitivity and specificity of Medicare race and ethnicity groupings.²⁵

Other Covariates

Demographic and clinical covariates were derived from Medicare files, including the denominator file and claims files. Demographic data included age at diagnosis and state of residence, with geographic regions based on the Census Divisions.²⁶ Disease-related and treatment-related variables included axillary lymph node involvement; axillary lymph node dissection; sentinel lymph node biopsy; receipt of any chemotherapy; receipt of doxorubicin or paclitaxel; any staging imaging; number of hospitalizations in the year after diagnosis; and number of medical oncology, radiation oncology, and surgery visits in the year after diagnosis. Variables indicating preventive healthcare and interactions with the healthcare system included mammography in the year before diagnosis and number of physician visits in the year before diagnosis. In addition, we calculated the severity of comorbid disease for each patient based on a modified Charlson comorbidity score validated in a prior claims-based study: 0 (no comorbidity), 1 (mild to moderate), or ≥2 (severe).²⁷ This score combined comorbidities recorded in Medicare claims during the 12 months before the patient’s cancer diagnosis. To enhance the specificity of comorbid disease diagnoses, patients must have had at least 1 inpatient (Part A) claim or at least 2 outpatient (Part B) claims more than 30 days apart.²⁷ Socioeconomic covariates were derived from the 2003 Area Resource File²⁸ linked to the Medicare files by county and state, and included rural/urban status, percentage of population (by county) living in poverty, median income, education level, density of surgeons, and density of radiation oncologists.

Statistical Analysis

All analyses were conducted using SAS software (version 9.1.3; SAS Institute Inc, Cary, NC), and all statistical tests assumed a 2-tailed a of 0.05. We calculated percentage RT use for the entire sample, by state and by region, and tested the unadjusted association between receipt of RT and race using the Pearson chi-square test. Bivariate associations between receipt of RT and other covariates were tested using the Pearson chi-square test for categoric variables and the Wilcoxon rank sum test for continuous variables. A multivariate logistic model tested the adjusted association between receipt of RT and race. Covariates were selected a priori based on significance in bivariate analyses (P < .25) and significance in prior studies of cancer patients.^19-24 Goodness of fit was assessed using the Hosmer and Lemeshow test. The final parsimonious multivariate logistic model was then also used to calculate adjusted rates of RT use.²⁹ A subsidiary model included socioeconomic covariates and tested the association in the subset of 33,172 patients who had socioeconomic covariate data available.

We further examined the association between receipt of RT and race in the subset of patients age <70 years (n = 7270), because prior evidence has suggested that the use of RT after BCS is more common in younger patients.⁵ In sensitivity analyses, we tested the association between race and receipt of RT in the subset of patients with a more restrictive (specific) definition of invasive breast cancer, which required the 2 diagnosis codes for invasive breast cancer to be ≥30 days apart (n = 33,114), and in a more restrictive definition of breast cancer, which excluded patients with any diagnosis code of ductal carcinoma in situ (DCIS) (ICD-9 diagnosis code 233.0) within 6 months of the invasive breast cancer diagnosis date (n = 22,301). We also tested a less restrictive treatment window of RT within 12 months of breast cancer diagnosis to take into account possible racial differences in treatment delays. Finally, in patients (of any age) who received RT, we also examined the association between completion of RT course and race.

RESULTS

Patient Characteristics

Of the 34,080 women with invasive breast cancer who were treated with BCS, 91% (N = 31,127) were white, 6% (N = 2077) were black, and 3% (N = 876) were another race. The mean age of the sample was 76 (standard deviation, 7) years. Approximately 73% of patients received RT after BCS, and 13% received chemotherapy as part of their initial treatment course. Approxmately 39% of patients had a claim indicating axillary lymph node dissection or axillary lymph node involvement, and 55% had a claim indicating sentinel lymph node biopsy (Table 2).

Table 2.

Sample Characteristics

Characteristic	Percentage
Demographic
Mean age (SD), y	76 (7)
Race
White	91
Black	6
Other	3
Region
West, Pacific West	11
West, Mountain West	5
Midwest, West North Central	7
Midwest, East North Central	19
Northeast, New England	7
Northeast, Mid-Atlantic	16
South, South Atlantic	21
South, West South Central	9
South, East South Central	5
Not known	1
Clinical
No comorbid diseases	32
Axillary lymph node involvement	39
Sentinel lymph node biopsy	55
Radiotherapy	73
Chemotherapy	13
Socioeconomic
Mean income (SD), $US	43,429 (10,649)
Education
Mean percent completed less than ninth grade (SD)	7 (4)
Mean percent completed college or higher (SD)	24 (9)
Mean surgeon density (per 100,000) (SD)	13 (10)
Mean radiation oncologist density (per 100,000) (SD)	1.5 (1.4)

SD indicates standard deviation.

Race and RT Use

Significant racial differences were found to exist in the receipt of RT; 74% of white women, 65% of black women, and 66% of other-race women received RT after BCS (P < .001). After adjustment for demographic, clinical, and socioeconomic covariates, white women were still significantly more likely than black women to have received RT after BCS (odds ratio [OR], 1.48; 95% confidence interval [95% CI], 1.34–1.63 [P < .001]). In addition, there was a trend toward white women also being more likely than other-race women to have received RT after BCS (OR, 1.22; 95% CI, 1.04–1.42 [P = .01]) (Table 3). Other variables found to be independently associated with higher rates of RT use included geographic region, younger age, lower comorbidity score, treatment with doxorubicin, axillary lymph node dissection, use of staging imaging, use of mammography before cancer diagnosis, and greater number of physician visits (interactions with the healthcare system) in the year before cancer diagnosis. In the subset of patients with complete socioeconomic covariate data available, the association between receipt of RT and race was unchanged. In this model, lower poverty level, higher median income, higher education level, higher surgeon density, and higher radiation oncologist density were also found to be associated with receipt of RT.

Table 3.

Racial Disparities in the Use of Radiotherapy After Breast-Conserving Surgery for Breast Cancer

Model: RT use	OR	95% CI	P
Unadjusted
White vs black patients	1.52	1.38–167	<.001
White vs other* patients	1.46	1.27–1.68	<.001
Adjusted†
White vs black patients	1.48	1.34–1.63	<.001
White vs other* patients	1.22	1.04–1.42	.01

^*RT indicates radiotherapy; OR, odds ratio; 95% CI, 95% confidence interval.

† Adjusted for age, comorbidity, chemotherapy (doxorubicin or paclitaxel), axillary lymph node involvement, staging, imaging, surgeon visits, mammography, physician visits, and region.

Sensitivity analyses using more restrictive definitions of invasive breast cancer did not appear to alter the magnitude or significance of the associations between race and RT use significantly. Sensitivity analyses using a less restrictive treatment window also did not appear to alter the magnitude or significance of these associations.

In the subset of patients age <70 years, 82% received RT after BCS. Specifically, 83% of white women, 73% of black women, and 78% of other-race women received RT after BCS (P < .001). On multivariate analysis of this younger group, white women were still found to be significantly more likely than black women to have received RT after BCS (OR, 1.73; 95% CI, 1.40– 2.14 [P = .003]), although there was no significant difference noted between white women and other-race women (OR, 1.28; 95% CI, 0.86–1.90 [P = .22]). The interaction term between race (black vs white race) and age (age <70 years vs ≥70 years) was not found to be statistically significant (P = .19).

In the subset of patients who received RT and had the number of treatments documented, 85% were found to have completed a course of therapy, whereas 15% had an incomplete course. Approximately 85% of white women, 85% of black women, and 82% of other-race women completed their RT course (P = .02). After adjusting for covariates, there remained no significant difference in rates of completion by race (black vs white women: OR, 1.00; 95% CI, 0.83–1.20 [P = .99] and other-race vs white women: OR, 1.02; 95% CI, 0.73–1.43 [P = .89]).

Geographic Variation in Racial Disparities

We observed significant geographic variations in racial disparities. The absolute difference in the percentage of RT use between white and black women by state ranged from −28% to 74%. However, for most states, the percentage of white women who received RT was greater than the percentage of black women who received RT, and in no state was the rate of RT use found to be significantly greater in black women compared with white women (Fig. 1). In areas of the northeastern and southern United States, the racial disparity between white and black women with regard to RT use was particularly pronounced (Fig. 2) (Table 4). Our data were insufficient to assess variations by state or region for other-race women.

Figure 1.

Percentage radiotherapy use in (a) white patients versus (b) black patients is shown. Gray shading indicates that the sample size was too small to provide meaningful data.

Figure 2.

Absolute difference in the rate of radiotherapy (RT) use between white and black patients is shown. An absolute difference <0% indicates that the percentage of black women receiving RT was greater than the percentage of white women. Gray shading indicates that the sample size was too small to provide meaningful data.

Table 4.

Unadjusted and Adjusted Regional Variations in the Rate of Radiotherapy Usage Between White and Black Patients With Breast Cancer

		Unadjusted			Adjusted*
Region	States	% Whites	% Blacks	P	% Whites	% Blacks	P
West, Pacific West	AK, CA, HI, OR, WA	72	54	<.001	74	61	.07
West, Mountain West	AZ, CO, ID, MT, NV, NM, UT, WY	77	78	.88	72	67	.59
Midwest, West North Central	IA, KS, MN, MO, NE, ND, SD	74	73	.94	74	70	.67
Midwest, East North Central	IL, IN, MI, OH, WI	76	72	.04	79	73	.15
Northeast, New England	CT, MA, NH, ME, RI, VT	72	69	.53	78	71	<.001
Northeast, Mid-Atlantic	NJ, NY, PA	71	57	<.001	70	62	.01
South, South Atlantic	DE, DC, FL, GA, MD, NC, SC, VA, WV	76	68	<.001	73	64	.009
South, West South Central	AR, LA, OK, TX	74	63	.001	69	59	<.001
South, East South Central	AL, KY, MS, TN	72	57	<.001	78	60	.001

^*Adjusted for all covariates except state/region.

DISCUSSION

In our analysis of a national sample of US women aged ≥65 years who were diagnosed with invasive breast cancer, white women (74%) were significantly more likely than black women (65%) to receive RT after BCS. The results of the current study also suggested that white patients were also more likely than patients of other races (66%) to receive RT after BCS. Furthermore, even after we took into consideration other covariates such as patient comorbidity, white women were still nearly 50% more likely than black women to receive RT. The racial disparity between whites and blacks persisted even among the younger patients in the cohort, those age <70 years, in whom RT after BCS was more common. Our analysis also identified geographic variations in the magnitude of treatment disparities between white and black patients across the United States, although some degree of treatment disparity existed in most states. Furthermore, the disparities across the nation persisted even after adjusting for variations in socioeconomic factors such as income, education, and access to healthcare resources. The large magnitude of racial disparities in RT use identified in the current study is concerning given that the administration of RT after BCS in patients with early invasive breast cancer reduces the risk of disease recurrence and breast cancer-related death and is considered the standard of care.1–4,30

Other studies have also described disparities in breast cancer care between black patients and white patients.¹⁵ In a recent secondary analysis of the NAFTA trial, in which all patients in the study cohort were treated with BCS and hormonal therapy, 80% of all enrollees received RT. Similar to the 9% to 10% absolute difference noted in the current study, there was a 12% absolute difference found between white and black patients in the rate of RT use in the NAFTA study.⁵ A recent analysis of the Surveillance, Epidemiology, and End Results (SEER)-Medicare cohort reported an 8% absolute difference between white and black patients with regard to the receipt of RT after BCS, with 86% of white women and 78% of black women between the ages of 66 and 85 years with invasive breast cancer receiving the treatment.³¹ A limitation of this study was that SEER-Medicare sampled ≤ 26% of the US Medicare population, compared with the current study, which sampled the Medicare population in its entirety.

A previous study of breast cancer patients who were treated in the state of Florida found that Hispanic women were significantly less likely to receive local therapy for breast cancer than non-Hispanic white women,¹⁴ whereas another study of patients in Hawaii found that Asian women may be less likely to receive therapy than white women.³² In our cross-sectional study, we were unable to explore the use of RT in specific nonwhite, nonblack ethnic groups because of the limited sample size of each ethnic group in our sample. Future analyses may seek to focus on exploring national patterns in potential treatment disparities affecting women of other racial/ethnic groups.

The underlying reasons for the racial disparities observed in the current study cohort remain to be determined.³³ Prior studies have identified access to care and socioeconomic factors (such as the patients’ income and education, the cost of RT, or the availability of supplemental insurance) as factors that may influence racial disparities in cancer care,^34,35 but even after adjustment for markers of healthcare access in the current analysis, such as number of physician visits and use of mammography before cancer diagnosis, racial disparities persisted. Other unmeasured potential explanatory factors include the impact of the physician-patient interaction.³⁶ For example, it is not known whether physicians offer treatment less frequently to nonwhite patients, whether substandard care occurs more frequently in predominantly nonwhite communities, or whether nonwhite patients are more likely to decline treatment. Additional social factors, such as culturally specific health beliefs, the presence of social support, and marital status,³⁷ also could affect racial disparities in care and may be important variables to explore in future studies seeking to identify specific barriers in breast cancer care.

The underlying causes of the geographic variations in racial disparities also require further study. Several published analyses have indicated that geographic variations in breast cancer care exist,^7,8,38 but to the best of our knowledge no prior study has described how geographic patterns in care may be modified by patient race. Our analysis found that disparities between white and black patients in RT after BCS appeared most pronounced in the northeastern and southern regions of the United States. Given that racial disparities do not appear to be concentrated in a single area of the country, the underlying reasons for disparities in the different regions of the United States are likely to be varied, complex, and multifactorial. For example, rural/urban differences may play a role in the geographic variations in treatment disparities. Prior studies have identified patients’ distance from RT facilities as a potential barrier to treatment,^12,39 and suggested that patients living in rural areas may have less access to RT.^39,40 However, a recent study reported similar rates of RT use after mastectomy and after lumpectomy in urban versus rural locales.⁴¹ The study’s results suggest that disadvantaged patients in either setting may experience distinct but persistent treatment barriers. Prospective efforts to overcome disparities in cancer treatment, such as the Patient Navigation program,⁴² are currently being evaluated and may serve as an important source for identifying the key barriers to cancer care in different geographic regions of the United States.

The current study has limitations to consider. Our sample focused on older breast cancer patients with continuous fee-for-service Medicare Parts A and B insurance coverage. Therefore, future studies will be required to validate the magnitude and significance of racial disparities in the treatment of younger patients. In addition, studies are also needed to determine the magnitude of racial disparities among patients with other insurance status, such as those with Medicaid, given that prior evidence suggests that patients with Medicare insurance coverage are actually more likely to receive appropriate breast cancer treatment.⁴³ Furthermore, our definitions of invasive breast cancer and RT were claims-based and, therefore, may be subject to misclassification bias. However, the results of the current study demonstrated reproducibility, even after applying the more restrictive criteria to exclude potential metastatic disease and DCIS. Prior studies indicate a high degree of validity for claims-based treatment variables and suggest that, although our treatment definition was unlikely to be 100% sensitive, our percentage use may underestimate true use rates by only approximately 2%.³⁷ Finally, the racial distribution of patients in the current study sample differs from the general population distribution. Notably, 6% of our sample was black, whereas 9% of US women (age >65 years) are black.²⁶ This difference is likely due to the tendency of nonwhite patients to receive mastectomy over BCS even for the treatment of early stage disease.^44–46

Conclusions

To the best of our knowledge, the current study provides 1 of the first available data sets describing a quality indicator for breast cancer care across a comprehensive national sample of older breast cancer patients and was able to identify substantial racial disparities in care across the United States. Our analysis helps define the scope of the treatment disparities in RT after BCS and underscores the concern that this treatment disparity occurs not merely in isolation but is instead a problem that exists on a national scale. Future efforts to improve breast cancer care will require identifying and overcoming the underlying causes of these racial disparities. As additional data become available, future studies may also explore changes in the magnitude of disparities over time and the effect of these disparities in care on breast cancer outcomes.