This retrospective cohort study used the Surveillance, Epidemiology, and End Results (SEER)-Medicare linked database to identify a diverse cohort of breast cancer patients. The SEER cancer file provided patient, tumor, and treatment information among patients diagnosed in the 18-state catchment area.(32) We used Medicare enrollment and claims files, including race and vital status information collected by the Social Security Administration, for 1999 through 2019. Fee-for-service (FFS) claims from inpatient and skilled nursing facilities (MedPAR), outpatient (OUTPAT), and professional services National Claims History (NCH) files were analyzed. We also examined Part D (prescription drug) events for patients with the corresponding coverage. This project was approved by the University of Iowa Institutional Review Board.

Women diagnosed with a first primary breast cancer between 2000 and 2017 were eligible for inclusion (Figure 1). The ‘Primary Cohort’ had malignant, invasive breast tumors (International Classification of Diseases for Oncology (ICD-O) behavior code 3).(33) Inflammatory carcinoma and Paget disease were excluded (ICD-O-3 histology 8530, 8540-8543). Patients of any stage at diagnosis – I-IV and those with unknown stage – were eligible. Patients diagnosed before age 66 or whose tumors were identified at time of death were excluded. At least 12 continuous months of FFS Medicare coverage prior to diagnosis was required to allow for evaluation of baseline comorbidities. Patients who died or whose FFS coverage lapsed within three months after diagnosis and those with other or unknown race were excluded. From the Primary Cohort, two distinct subsets of patients were identified: 1) Part D eligibile, defined as having three or more months of continuous Part D coverage after diagnosis; 2) those diagnosed in 2010 or later, when human epidermal growth factor receptor 2 (HER2) status was first reported in SEER.

Rate ratios (RRs) for each component of care by race-ethnicity were estimated using log-linear Poisson regression models. Rates were calculated as a function of person-time (months) of follow-up for outcomes during the treatment period by including an offset in the Poisson models. NHW patients were selected as the reference group for RRs because they were the largest race-ethnicity group. Crude and adjusted RRs with corresponding 95% confidence intervals (CIs) were estimated with robust variance estimators. All analyses were also performed stratified by stage at diagnosis (I-III, IV, and unknown/missing) and age at diagnosis (66-69, 70-75, 76-80, 81+), shown in Supplementary Tables 2–3. Patient demographic and health status covariates were adjusted for in all models and analyses of first-line treatment controlled for stage at diagnosis and HR status, if not already accounted for by subgroup selection or stratification. To assess the importance of demographics – a mediator of the relationship between race-ethnicity and receipt of care – to observed associations, we performed sensitivity analyses adjusted for all covariates except marital status and low-income subsidy (Supplementary Table 4).

Two-tailed tests with a significance level of 0.05 were utilized for all analyses. All analyses were performed in SAS software (v9.4; RRID:SCR_008567). Data visualization was performed using R (v4.2.1; RRID:SCR_001905) with the ggplot2 package (RRID:SCR_014601).

SEER-Medicare data are available upon NCI approval of a specific research project; more information can be found at https://healthcaredelivery.cancer.gov/seermedicare/obtain/.

Approximately 90% of the Primary Cohort had a claim for diagnostic mammography (Table 2). Most patients (82.9%) had the guideline-preferred initial breast biopsy performed percutaneously.

NHW women had the highest percentage of patients receiving diagnostic mammography and API had the highest proportion receiving an initial needle breast biopsy (Figure 2A). Conversely, Black and AIAN women had the lowest proportions of patients to receive diagnostic mammography and initial needle breast biopsy. All race-ethnicity groups had statistically significantly lower rates of diagnostic mammogram compared to NHW women (Table 3). However, only Black and AIAN women had significantly lower rates of both diagnostic mammogram and initial needle breast biopsy compared to NHW. In stratified analyses, Black and AIAN women had significantly lower rates of diagnostics primarily among women with non-metastatic (stage 1-3) disease and those diagnosed at younger ages (66-69 years; Supplementary Tables 2–3)).

Over 80% of the Primary Cohort received a lymph node biopsy (Table 2). Stage at diagnosis and tumor grade were documented for 93.2% and 90.5%, respectively, of the Primary Cohort. HR status was determined for 90% of the Primary Cohort and HER2 status was determined for 92.0% of patients with the HER2 variable available.

Black women had the lowest proportion of patients with known grade at diagnosis, to receive any lymph node biopsy, and to have HR status determined (Figure 2B). RRs for Black women compared to NHW were significantly lower for these three outcomes before and after adjustments (Table 3). The difference in known grade between Black and NHW women was largest among patients with unknown stage at diagnosis; effects were statistically significant for unknown and stage 1-3 patients and across all age groups (Supplementary Tables 2–3).

AIAN women had the lowest proportions with known stage at diagnosis or documented HR status (Figure 2B). AIAN women were significantly less likely to have known stage than NHW (Table 3). Among patients diagnosed at stage 4, AIAN women had notably lower rates of HR and HER2 status determination than NHW (RR_{HR, adjusted} (95% CI) 0.85 (0.7-1.0); RR_{HER2, adjusted} 0.72 (0.52-0.99); Supplementary Table 2).

API patients had the highest proportion of patients receiving all clinical workup and staging procedures (Figure 2B). API women were significantly more likely to have known grade and HR status than NHW women before and after adjustments (Table 3). Stage-stratified analyses revealed statistically significant increased rate of node biopsy for API versus NHW women among the stratum of stage 4 patients (Supplementary Table 2).

Over 90% of the Primary Cohort had cancer-directed surgery (91.6%); 49.2% received breast conserving surgery (BCS) and 42.5% received mastectomy (Figure 3A). Radiation, which is guideline-recommended for all patients who receive BCS, as well as patients who receive mastectomy and have N2+ or T3 disease, was provided to 78.3% of BCS-treated patients and 60.9% of mastectomy-treated patients eligible for radiation. Radiation is not guideline-recommended for patients who do not receive surgery but was received by 21.8% of these patients. Chemotherapy is recommended for all women with triple negative tumors and 59.6% of these patients received it. HER2-targeted therapy was observed in 59.7% of HER2+ patients. Most (79.5%) HR+ patients with Part D eligibility received at least one dose of hormone therapy.

Black women had the lowest proportion of patients to receive cancer-directed surgery (Figure 3A). Black women were significantly less likely to receive any cancer-directed surgery compared to NHW after adjusting for patient and tumor characteristics (Table 3). The magnitude of the disparity was largest among patients with unknown stage and aged 81 or older at diagnosis (Supplementary Tables 2–3). AIAN women had the second-lowest proportion to receive surgery of any race-ethnicity group; RRs for surgery compared to NHW were not statistically significant after adjustment for clinical and demographic characteristics.

Among women who received BCS, AIAN and Black women had the lowest proportions to receive radiation (Figure 3B). Both AIAN and Black women had 0.99 times the rate of radiation as NHW among BCS-treated women, however effects for AIAN women were not statistically significant (Table 3). Conversely, API and HW women had significantly higher rates of radiation than NHW among BCS-treated patients. Of all mastectomy-treated patients eligible for radiation, AIAN and API women had the highest proportions to receive radiation therapy and were significantly more likely to receive radiation relative to NHW women. There were no significant differences in the rate of radiation therapy among patients who did not receive any surgery.

Among triple negative patients, HW and Black women had the highest proportion of patients receiving chemotherapy, while NHW had the lowest (Figure 3C). HW and Black women were significantly more likely to initiate chemotherapy compared to NHW before and after adjustment for patient and tumor characteristics (Table 3). Black women had significantly higher rates of chemotherapy among stage 1-3 and 4 and among patients diagnosed at the youngest and oldest ages (Supplementary Tables 2–3). HW women only had significantly higher rates among stage 4, compared to NHW (Supplementary Table 2).

Among HER2+ patients, AIAN and Black women had the lowest proportions of patients to receive HER2-targeted therapy while API women had the highest initiation proportion (Figure 2E). There were no statistically significant differences in HER2 therapy initiation for AIAN or Black women compared to NHW, however the magnitude of effect may reflect clinically significant differences (RR_{AIAN, adjusted} 0.82 (0.62-1.08); RR_{Black, adjusted} 0.95 (0.90-1.01)).

AIAN women had the lowest proportion of patients to receive hormone therapy among HR+ patients with Part D coverage while API women had the highest (Figure 3C). Additionally, AIAN women were the only group with significantly lower rates of hormone therapy initiation than NHW before and after adjustments (RR_adjusted 0.92 (0.85-0.99); Table 3). The disparity in hormone therapy initiation was largest among unknown and stage 4 patients, but neither were statistically significant (RR_{Stage Unknown, adjusted} 0.81 (0.50-1.32); RR_{Stage 4, adjusted,} 0.85 (0.53-1.35); Supplementary Table 2). Among stage 1-3 patients, AIAN women had lower rate of hormone therapy initiation, but the effect was only significant after adjustments (RR_adjusted 0.93 (0.86-0.99)). API had significantly higher rates of hormone therapy initiation compared to NHW before and after adjustments in the overall analysis and among stage 1-3 patients.

RRs from models adjusted for all covariates except marital status and low-income subsidy tended to be farther from the null (i.e., magnitude of the difference was larger) than primary models (Supplementary Table 4). API and HW women were significantly less likely to receive an initial needle biopsy compared to NHW in the sensitivity analysis, however the disparity was still larger for Black and AIAN women. Without adjustment for demographics, AIAN women were significantly less likely to receive a node biopsy and to have HR status documented than NHW (RR_{Node biopsy} 0.96 (0.94-0.99); RR_{HR documentation} 0.98 (0.95-0.99)). Black women had significantly lower rates of known stage at diagnosis and HER2-targeted therapy compared to NHW (RR_Stage 0.992 (0.987-0.996); RR_HER2 0.93 (0.88-0.99)). There were no significant differences between HW and NHW for any primary treatment modality when marital status and low-income subsidy were not adjusted for.

Due to the nature of secondary data analyses, our study is prone to some limitations that warrant consideration in interpretation of the findings. Selection bias may result from differences in age at diagnosis and life expectancies between race-ethnicity groups. Rather than excluding patients with missing values of any variable, we treated unknown as its own value, to mitigate selection bias from demonstrated racial-ethnic patterns of missing data.(45, 62) Interpretation of unknown values requires the assumption that this information would also be unknown to the patients’ providers. We believe this to be reasonable based on the completeness of the SEER-Medicare database, however this may not be correct in all cases.(62) Over one-third of the initial population was excluded due to lack of continuous Medicare coverage in the year prior to diagnosis; of note – this population may be more vulnerable to disparities in care than the eligible study population with continuous coverage. Survival bias may have been introduced by requiring patients lived at least 3 months post-diagnosis. However, this was necessary to provide adequate time for patients to receive some cancer care services. The number of analyses performed increases the chance of finding a significant effect as an artifact of multiple testing, however the potential implications of finding a spurious association outweigh the risks of not doing this work.(63) Moreover, while some comparisons were based on small sample sizes, particularly among AIAN patients in stage-stratified analyses, it remains important to conduct analyses with/for small populations despite wide confidence intervals.(64)

This study has several strengths. This is the first study to examine differences in treatment within the context of earlier components of care that may contribute to differences in treatment later in the continuum of care. We used multiple sources of data to maximize the accuracy of our race-ethnicity characterization. We imposed minimal cohort criteria to maximize generalizability to the broader population of older women with breast cancer in the United States. Our measures of evidence-based care were identified from multiple clinical practice guidelines and quality metrics, maximizing the relevance of these findings to routine clinical settings. Rate ratios are the preferred comparative effect estimate for cohort studies and common outcomes.(65–67) Additionally, the RR accounted for each patients’ duration of follow-up, which allowed us to minimize eligibility criteria while still appropriately accounting for the amount of time in which the outcome could be observed.