Fax: (301) 435–3710
Differences in the quality of breast cancer care among vulnerable populations
Article first published online: 6 OCT 2005
Copyright © 2005 American Cancer Society
Volume 104, Issue 11, pages 2347–2358, 1 December 2005
How to Cite
Haggstrom, D. A., Quale, C. and Smith-Bindman, R. (2005), Differences in the quality of breast cancer care among vulnerable populations. Cancer, 104: 2347–2358. doi: 10.1002/cncr.21443
- Issue published online: 18 NOV 2005
- Article first published online: 6 OCT 2005
- Manuscript Accepted: 28 JUN 2005
- Manuscript Revised: 1 JUN 2005
- Manuscript Received: 26 APR 2005
- University of California, San Francisco, Division of General Internal Medicine, General Internal Medicine Clinical Research Fellowship
- Health Resources and Services Administration Faculty Development Grant. Grant Number: 1D14 HP001 78–01
- California Breast Cancer Research Program. Grant Number: 9PB-0205
- Department of Defense. Grant Number: DAMD17–99–1-9112
- National Institutes of Health (NIH). Grant Number: CA86032
- breast cancer;
- African Americans;
- Hispanic Americans;
- Asian Americans;
- physician's practice patterns;
- women's health
It is unknown whether differences in the quality of breast cancer care among women from racial and ethnic minority groups, the elderly, and rural areas have changed over time across the continuum of care.
The linked Surveillance, Epidemiology, and End Results-Medicare database identified 22,701 women ages 66–79 years diagnosed with early stage breast cancer from 1992–1999. Multiple breast cancer processes of care were measured, including breast-conserving surgery, radiation therapy, documentation of estrogen receptor status, surveillance mammography, and a combined measure of “adequate care”.
African-American and Hispanic women were significantly less likely to receive adequate care than White women in unadjusted comparisons (54.7% and 58.0% vs. 68.4% for African-American and Hispanic vs. White women) and adjusted comparisons (adjusted odds ratio [AOR] 0.67; 95% confidence interval [95% CI] 0.59–0.76, and AOR 0.77; 95% CI 0.66–0.90 for African-American and Hispanic women, respectively). The proportion of Asian/Pacific Islander women receiving adequate care was similar to White women. When considering only women diagnosed with breast cancer from 1997–1999, African-American women remained less likely than White women to receive adequate care (AOR 0.63; 95% CI 0.50–0.79). Women ages 75–79 years were less likely to receive adequate care compared with women ages 66–69 years (AOR 0.74; 95% CI 0.69–0.80), and women from rural (vs. metropolitan) areas were less likely to receive adequate care (AOR 0.81; 95% CI 0.73–0.89).
The quality of breast cancer care is lower among vulnerable populations across the continuum of care, and many of these differences have not improved in more recent years. Cancer 2005. © 2005 American Cancer Society.
Substantial racial and ethnic differences in breast cancer outcomes and the use of healthcare services have been observed. African-American women have a lower incidence of breast cancer, but death rates among African-American women are higher than White women, and the mortality gap has increased.1 African-American women have been less likely to receive radiation therapy (XRT) compared with White women,2, 3 a treatment difference that may partly explain these differences in breast cancer mortality. Hispanic and Asian/Pacific Islander women in the United States both have been less likely to receive breast-conserving surgery (BCS) surgery,4, 5 a treatment difference that may adversely impact their quality of life (QOL). Other vulnerable populations are also more likely to receive inadequate care for breast cancer, including the elderly, women who live in rural areas, and women of lower socioeconomic status.6–8
To understand the reasons for differences in cancer outcomes, it is important to assess the quality of breast cancer care. When assessing the quality of breast cancer care, process measurements have the advantage of occurring more frequently than mortality and of being potentially modifiable within the context of the healthcare delivery system.9 The Institute of Medicine (IOM) report, Ensuring Quality Cancer Care,10 stressed the importance of delivering high-quality services across the full span of the cancer care continuum. We sought to evaluate differences in breast cancer treatment and surveillance as a way to characterize differences in the quality of breast cancer care across as broad a spectrum of healthcare services as possible.
Several evidence-based guidelines regarding appropriate treatment processes for early stage breast cancer have been published that enable evaluation of the quality of breast cancer care. Breast-conserving treatment (including both BCS and XRT) was recommended by the 1990 National Institutes of Health Consensus Panel as preferable to mastectomy because of equivalent survival and improved QOL.11 BCS without XRT is considered inadequate because BCS is associated with a higher cancer recurrence rate without radiation. Because assessment of estrogen receptor (ER) status is necessary to determine whether chemotherapy is appropriate, the adequate documentation of ER status can also be used to measure quality. Finally, the quality of care provided in the survivorship phase of cancer is important, as many cancer patients live for years after initial treatment. Surveillance mammography on an annual basis after diagnosis with breast cancer is recommended as a clinical guideline.12
By using part of the framework implemented in the National Healthcare Disparities Report, we set out to describe differences in the use of breast cancer care by vulnerable “priority” populations.13 These populations include women from racial and ethnic minorities, the elderly, rural populations, and low-income groups. We planned to expand and update previous knowledge regarding differences in the quality of breast cancer care among these populations using the most recent information available. We used population-based data from the Surveillance, Epidemiology, and End Results (SEER)-Medicare program that includes breast cancers diagnosed from 1992 to 1999 and treated through 2002.
MATERIALS AND METHODS
The SEER-Medicare database used for this analysis was created as a collaborative effort of the National Cancer Institute, SEER program, and the Centers for Medicare and Medicaid Services (CMS) to create a population-based source of information for cancer-related research. The SEER cancer registries identified all incident breast cancer diagnoses in selected areas from 1992 to 1999 and provided diagnostic and treatment information within 4 months after the date of diagnosis. The Medicare files provided claims information about breast cancer treatment (including primary breast cancer treatment that may not have been captured in the SEER data) and surveillance after diagnosis through 2002. The process for linking these data has been described by Warren et al.14 The SEER program includes population-based tumor registries in several geographic areas and covers approximately 14% of the U.S. population.
Women ages 66 to 79 years who were diagnosed with Stage I or II breast cancer from 1992 to 1999 were included in our study. We included women beginning at age 66 years so that there would be at least 1 year of Medicare claims before diagnosis by which to assess comorbidity. We excluded women older than age 79 years because of more uncertainty regarding appropriate care among these women, and we also excluded women with tumors > 5 cm because neoadjuvant chemotherapy, as opposed to surgery, is commonly the first course of treatment for these women.
Women were excluded who were diagnosed with another cancer before, or within 2 years after, the incident breast cancer to ensure all treatment was intended for the identified breast cancer. Medicare does not receive individual billing claims for patients enrolled in HMO plans, and healthcare coverage may be provided as part of either Medicare Part A or B. Therefore, women were excluded who had HMO coverage or were not covered by both Parts A and B of Medicare 1 year before, or 2 years after, diagnosis with breast cancer. Finally, we excluded those women who died during the period when they were eligible for treatment.
Race or ethnicity was defined as non-Hispanic White, African-American, non-Hispanic; Hispanic, Asian/Pacific Islander, Native American, or other/unknown using the recoded SEER race variable.15 Hispanic women were identified by SEER through an algorithm that uses Spanish surnames. Other categories of race/ethnicity were determined by SEER from medical records and registration information. The Asian/Pacific Islander category includes Chinese, Japanese, Filipino, and Hawaiian women.
Age was calculated from the patient's birth date using SEER and then grouped into 5-year categories. Rural residence was obtained from SEER and dichotomized by designating the assigned area as rural (county either distant or adjacent to a metropolitan area, and including counties with a population of ≤ 20,000) or metropolitan (county in a metropolitan area, and including areas with a population of ≥ 250,000). Socioeconomic status was obtained from SEER by using median income in the patient's census tract; if census tract information was missing, then median income in the patient's zip code was used.15
Individual year of diagnosis, SEER region, tumor size, and American Joint Committee on Cancer tumor stage were obtained from SEER and included as categoric predictors of cancer treatment. Comorbidity was measured with a Charlson-based comorbidity index derived from Medicare inpatient and outpatient claims. This index has previously been shown to be predictive of 2-year mortality and the receipt of less aggressive therapy, specifically BCS without XRT.16
Processes of Care
Surgery receipt was measured using information from both SEER and Medicare.2, 17–19 Surgery was considered to occur if it was noted in either SEER or Medicare (Table 1). In cases of disagreement concerning type of surgery, the most invasive surgery (mastectomy) noted in either SEER or Medicare was used. Radiation therapy receipt was measured using information from both SEER and Medicare20, 21 (Table 1). If XRT was noted to occur in either SEER or Medicare, then it was considered to have occurred. Treatment with both surgery and XRT were measured within a 6-month interval after breast cancer diagnosis.
|Breast cancer process of care||SEER||Medicare|
|Breast-conserving surgery||segmental/subtotal mastectomy, lumpectomy, quandrantectomy, tylectomy, wedge resection, nipple resection, excisional biopsy, or partial mastectomy, not otherwise specified||ICD-9-CM procedure codes: 85.20-85.23|
|CPT codes: 19120, 19125, 19126, 19160, 19162|
|Mastectomy||subcutaneous, total (simple), modified radical, radical, or extended radical mastectomy||ICD-9-CM procedure codes: 85.41, 854.2, 854.3, 85.44, 85.45-85.48|
|CPT codes: 19240, 19220, 19180, 19182|
|Radiation therapy||Beam radiation, radioactive implants, radioisotopes, or other radiation||ICD-9-CM procedure codes: 92.21-92.29|
|Revenue center codes: 0330, 0333|
|CPT codes: 77401-77499, 77750-77799|
|Estrogen receptor status||Data from SEER|
|Surveillance mammography||CPT codes: 70690-76092|
Estrogen receptor status documentation was measured using SEER data alone, as this information is not available in Medicare. ER status was categorized as 1) positive, 2) negative, 3) borderline, 4) not done, 5) ordered but no result recorded in chart, and 6) unknown or not documented in patient records. These categories were collapsed and dichotomized into adequate (Categories 1–3) and inadequate (Categories 4–6) ER status documentation.
Mammography surveillance was measured by using information from Medicare alone. For each patient, mammograms that occurred within the first 6 months after initial surgical treatment were considered part of the treatment period, and mammograms that occurred 7–18 months after initial surgical treatment were considered to have occurred for surveillance.22 All mammograms performed within the first 3–5 years after a breast cancer diagnosis are labeled as diagnostic, and therefore, we included all billing codes that may have been used to identify a mammogram (Table 1).
Adequate care was defined as either the receipt of XRT among women receiving BCS, or the receipt of mastectomy. In addition, adequate care was defined as adequate documentation of ER status and the receipt of surveillance mammography among all women. Therefore, a woman was defined as receiving adequate care if she received all of the following: either mastectomy or BCS with XRT, adequate documentation of ER status, and surveillance mammography.
We calculated the proportion of women receiving BCS among women with Stage I or II breast cancer. XRT is recommended for women undergoing breast-conserving treatment, and thus, we calculated the proportion of women receiving XRT only among women who had received BCS. Both the proportion of women with adequate documentation of ER status and surveillance mammography were calculated among all women with Stage I or II breast cancer who underwent surgery. The proportion of women with adequate care was calculated among all women with Stage I or II breast cancer. Changes over time in each breast cancer process of care among different racial and ethnic groups were charted with moving 3-year averages. Changes over time by year in the overall proportion of women receiving each process of care were evaluated for statistical significance using a nonparametric test for trend across ordered groups.23
We evaluated the associations among race/ethnicity, age, and rural status, and the receipt of each breast cancer process of care (BCS, XRT, adequate documentation of ER status, mammography surveillance) and adequate care using chi-squared tests. By using all of the predictors and covariates previously described, we created multivariate logistic regression models for the outcomes of receipt of BCS, XRT, and adequate care. Similar logistic regression models were developed for the outcomes of adequate documentation of ER status and surveillance mammography, but these models excluded clinical predictors of tumor size and stage, as women should receive these care processes regardless of tumor size or stage.
All multivariate logistic regression models were repeated, including median area level income as a predictor to determine whether socioeconomic status affects racial/ethnic, age, or rural differences in breast cancer care. Income was modeled as a continuous log term. Potential interactions between race/ethnicity and age, as well as race/ethnicity and income, were tested; there were no significant interactions in either case, and therefore, no interaction terms were included in the final models. Finally, all logistic regression models were repeated and limited to breast cancer diagnoses from 1997 to 1999 for the purpose of determining if any observed patterns of care persisted at the end of the decade.
There were 22,701 women ages 66–79 years diagnosed with early stage breast cancer from 1992 to 1999, including, 1137 African-American, 727 Hispanic, and 772 Asian/Pacific Islander women (Table 2). The majority of women (83.1%) lived in metropolitan areas, and most were diagnosed with Stage I tumors (65.8%) and tumors <2 cm in size (68.9%).
|No. of women||Percentage|
|Age in yrs|
|AJCC tumor stage|
|< 2 cm||15634||68.9|
|mean (SD)||$54,685 ($25,203)|
|Year of diagnosis|
|1992–1999 cases/yr||range 2,722–3,092/year|
Almost half (48.8%) of women received BCS, and among women who received BCS, 88.1% received XRT (Table 3). Among women who did not receive BCS, 45 (0.4%) had no surgery identified. Greater than 80% of women had adequate documentation of ER status or surveillance mammography after breast cancer surgery. By using a composite measure to aggregate individual processes of care, approximately two-thirds (67.3%) of women received adequate care.
|Breast-conserving surgery||Radiation therapy||Adequate ER documentation||Surveillance mammography||Adequate care|
|No. of women eligible||22,701||11,086||22,656||22,656||22,701|
|%||P value||%||P value||%||P value||%||P value||%||P value|
|White (reference group)||49.1||88.9||82.7||86.0||68.4|
|African-American||49.3||0.93||79.9||< 0.001||75.8||< 0.001||78.7||< 0.001||54.7||< 0.001|
|Hispanic||46.1||0.11||84.1||0.008||76.0||< 0.001||79.3||< 0.001||58.0||< 0.001|
|Age in yrs|
|66–69 (reference group)||49.9||91.9||82.2||87.2||69.7|
|75–79||47.5||0.005||82.0||< 0.001||82.5||0.67||82.5||< 0.001||63.7||< 0.001|
|Rural||35.1||< 0.001||85.9||0.009||85.5||< 0.001||85.9||0.22||71.3||< 0.001|
Race and Ethnicity
Asian/Pacific Islander women were significantly less likely to receive BCS compared with White women in both unadjusted (43.0% vs. 49.1%, Table 3) and adjusted comparisons (OR 0.52; 95% CI 0.43–0.63; Table 4). Hispanic women were not significantly less likely to receive BCS than White women in unadjusted comparison (46.1% vs. 49.1%, P = 0.11), but, in adjusted comparison (OR 0.85; 95% CI 0.72–0.99), this difference became statistically significant (P = 0.046). African-American women were no less likely to receive BCS than White women.
|Breast-conserving surgery||Radiation therapy||Adequate ER documentation||Surveillance mammography||Adequate care|
|AORa||95% CI||AORa||95% CI||AORb||95% CI||AORb||95% CI||AORa||95% CI|
|Age in yrs|
Among women who received BCS, African-American and Hispanic women were significantly less likely to receive XRT compared with White women in unadjusted (79.9% and 84.1% vs. 88.9% for African-American and Hispanic vs. White women) and adjusted comparisons (OR 0.55; 95% CI 0.44–0.70, and OR 0.68; 95% CI 0.50–0.94 for African-American and Hispanic women, respectively). Asian/Pacific Islander women were no less likely to receive XRT than White women.
Estrogen Receptor Status
Hispanic women were significantly less likely to have adequate documentation of ER status compared with White women in unadjusted (76.0% vs. 82.7%) and adjusted comparisons (OR 0.82; 95% CI 0.69–0.99). Similarly, African-American women were less likely to have adequate documentation of ER status than White women in unadjusted (75.8% vs. 82.7%) and adjusted comparisons (OR 0.87; 95% CI 0.75–1.00), although for the adjusted comparison, this difference bordered on statistical significance (P = 0.057). Asian/Pacific Islander women were more likely to have adequate documentation of ER status than White women in unadjusted comparison (86.5% vs. 82.7%), but this difference was no longer present in adjusted comparison (OR 0.87; 95% CI 0.67–1.11).
African-American and Hispanic women were significantly less likely to receive surveillance mammography than White women in unadjusted (78.7% and 79.3% vs. 86.0% for African-American and Hispanic vs. White women) and adjusted comparisons (OR 0.58; 95% CI 0.50–0.68, and OR 0.73; 95% CI 0.60–0.88 for African-American and Hispanic women, respectively). Asian/Pacific Islander women were no less likely than White women to receive surveillance mammography.
African-American and Hispanic women were significantly less likely to receive adequate care than White women in unadjusted (54.7% and 58.0% vs. 68.4% for African-American and Hispanic vs. White women) and adjusted comparisons (OR 0.67; 95% CI 0.59–0.76, and OR 0.77; 95% CI 0.66–0.90) for African-American and Hispanic women, respectively). Asian/Pacific Islander women were no less likely than White women to receive adequate care.
The receipt of most breast cancer processes of care declined with increasing patient age. Comparing women ages 75–79 years to women ages 66–69 years, there was a modest difference in receipt of BCS (OR 0.85; 95% CI 0.79–0.92, Table 4). Women aged 75–79 years were substantially less likely to receive XRT than women aged 66–69 years (OR 0.40; 95% CI 0.34–0.47). There was no significant difference in adequate documentation of ER status, but there was a significant difference in surveillance mammography between women ages 75–79 years and those ages 66–69 years (OR 0.67; 95% CI 0.61–0.74). Overall, the receipt of adequate care was lower among older (75–79 yrs) than younger (66–69 yrs) women (OR 0.74; 95% CI 0.69–0.80).
Women from rural areas were significantly less likely than women from metropolitan areas to receive BCS (OR 0.72; 95% CI 0.65–0.79, Table 4), XRT (OR 0.75; 95% CI 0.60–0.92), and adequate documentation of ER status (OR 0.67; 95% CI 0.59–0.77). Rural women were no less likely than metropolitan women to receive surveillance mammography. Overall, rural women were less likely to receive adequate care than metropolitan women (OR 0.81; 95% CI 0.73–0.89).
Including median area level income in the multivariate regression models did not significantly alter results among African-American women (results not shown in table). African-American women remained significantly less likely than White women to receive XRT (OR 0.62; 95% CI 0.48–0.79), surveillance mammography (OR 0.65; 95% CI 0.55–0.77), and adequate care (OR 0.72; 95% CI 0.63–0.82). Among Hispanic women, adjusting for median area level income attenuated the results. Hispanic women were no longer significantly less likely than White women to receive BCS (OR 0.93; 95% CI 0.79–1.10), XRT (OR 0.74; 95% CI 0.53–1.03), or adequate documentation of ER status (OR 0.84; 95% CI 0.70–1.01). Yet Hispanic women remained significantly less likely to receive surveillance mammography (OR 0.77; 95% CI 0.63–0.93) and overall adequate care than White women (OR 0.81; 95% CI 0.69–0.95). Lower use of breast cancer processes of care among older women and rural women also persisted after adjusting for income.
When we considered median area level income directly, we found that women living in areas of higher socioeconomic status were more likely to receive most processes of care. For every 25% increase in income, women had 7% greater odds of receiving BCS, 5% greater odds of receiving XRT, 4% greater odds of surveillance mammography, and 3% greater odds of receiving adequate care. There was no greater likelihood of adequate documentation of ER status with increasing income.
The proportion of women who received every breast cancer process of care, as well as adequate care, increased over time between the periods of 1992–1994 and 1997–1999 (Fig. 1). The receipt of BCS increased from 40.8% to 56.5%. The receipt of XRT (among women who received BCS) increased from 86.1% to 89.0%, the adequate documentation of ER status increased from 79.3% to 85.1%, and surveillance mammography increased from 82.8% to 88.4%. The proportion of women who received adequate care increased from 63.0% to 72.1%. The change over time by year was significant for every breast cancer process of care measured (P < 0.001).
Recent Breast Cancer Diagnoses
Many important differences among vulnerable populations persisted at the end of the decade. All differences between African-American and White women in breast cancer care remained significant when considering only breast cancers diagnosed from 1997 to 1999 (Table 5). In addition, differences in the adequate documentation of ER status between African-American and White women became statistically significant in adjusted comparison only when considering more recent cancer diagnoses (OR 0.70; 95% CI 0.54–0.91). However, there were no longer significant differences between Hispanic and White women in adequate care, or any individual process of care, when considering only more recent breast cancer diagnoses. As before, there was no significant difference in adequate care between Asian/Pacific Islander and White women diagnosed with breast cancer from 1997 to 1999. Among women from rural areas, lower use of individual processes of care was unchanged, although overall, rural women were no longer significantly less likely to receive adequate care. Age-related patterns of care remained largely the same when only more recent years were considered.
|Breast-conserving surgery||Radiation therapy||Adequate ER documentation||Surveillance mammography||Adequate care|
|No. of women eligible||8319||4678||8300||8300||8319|
|AORa||95% CI||AORa||95% CI||AORb||95% CI||AORb||95% CI||AORa||95% CI|
|Age in yrs|
Breast cancer care across the continuum of cancer treatment and surveillance outlined here has improved over time. Yet despite these improvements, only 67.3% of women received adequate care. There also remain significant gaps in the quality of care among vulnerable populations. African-American women were significantly less likely to receive adequate care compared with White women, and this gap persisted or worsened among women diagnosed with breast cancer later in time (1997–99). Gaps in quality of care among the elderly and women from rural areas and areas of lower socioeconomic status also persisted at the end of the decade.
Racial and Ethnic Differences
Asian/Pacific Islander women were less likely to receive BCS, but they were no less likely to receive any other breast cancer process of care. Lower rates of BCS have been observed previously among Asian/Pacific Islander women in California,4, 5 but the Asian/Pacific Islander population of our study is notably different because it includes women from other regions, including Hawaii. Hispanic women were also less likely to receive BCS; a similar pattern of lower use was observed previously in California.4
African-American and Hispanic women were less likely to receive XRT, although this difference did not persist among Hispanic women at the end of the decade. These racial and ethnic differences in XRT are consistent with a previous study that used data from SEER registries alone.24 Differences in receipt of XRT among women with early stage breast cancer is important because XRT appears to have a mortality benefit,25 although adequate numbers of older women have not been studied.26 A recent study found that African-American women had higher mortality after breast cancer than White women after adjustment for prognostic factors,27 and we would suggest that racial differences in treatment may contribute to these mortality differences.
African-American and Hispanic women were less likely to receive both adequate documentation of ER status and surveillance mammography. To our knowledge, previous studies have not addressed racial or ethnic differences in the documentation of ER status. A racial difference in mammography after surgery for early stage breast cancer was previously observed in 1991.28
When we considered only women diagnosed with breast cancer from 1997 to 1999, African-American women remained less likely than White women to receive adequate care and became significantly less likely to have adequate documentation of ER status. When we considered more recent diagnoses among Hispanic women, however, differences in care were no longer present, which suggests that breast cancer care may have improved among this population during the course of the decade.
Age-related patterns of care can be difficult to interpret in terms of appropriateness because of the low proportion of women older than age 70 years who enroll in clinical trials. Nonetheless, XRT is currently considered to be a necessary component of breast-conserving treatment,11 and we found that XRT decreased with increasing age. Declining use of XRT with advancing age has been observed previously and associated with patient preferences.29 Potential exists for both undertreatment among healthy older women, as well as overtreatment among the frail elderly.
Women who live in rural areas were less likely to receive BCS and XRT. These findings are consistent with a previous study7 and likely reflect multiple factors, including barriers to travel in rural areas that make access to XRT more difficult. In contrast, rural differences in surveillance mammography were not significant. Mammography examinations may not be influenced by the same geographic factors as more intensive therapies.
Previous research has suggested that racial differences in breast cancer treatment may be explained by socioeconomic status.8, 30 In our study, racial differences in breast cancer care did not change substantively when adjusted for median area level income. These findings likely differ from others because our study included multiple population-based cancer registries8 and measured a different spectrum of breast cancer processes of care.30 Hispanic women were no longer significantly less likely to receive many processes of care when we adjusted for income, which suggests that differences in socioeconomic status or social class31 may play a larger role in treatment differences among this population.
Lower median area level income was independently associated with lower use of most breast cancer processes of care. Although culture is commonly associated with race or ethnicity, there may also be a “culture” of poverty—denoting shared beliefs and behaviors—that influences patients' choices of specific treatments.32 Because median income was measured at the level of the census tract or zip code, these findings suggest that the socioeconomic status of the community in which a person lives is related to the quality of healthcare delivered to individuals in that community.
Explanations for Differences
Potential explanations for differences in patterns of care among vulnerable populations include poor healthcare access, regional variations, or a higher disease burden among these populations. Poor healthcare access, as measured by insurance coverage, was not clearly evident here given that all women were covered by Medicare Parts A and B. Differences in cancer care between similarly insured African-American and White patients have been observed previously in both breast cancer3 and lung cancer.33 Another dimension of access to consider in future research that uses Medicare claims is any unmeasured variation in supplemental insurance. In addition, the uninsured represent another vulnerable population that should be considered in future studies. Whereas significant regional variation in breast cancer care may be present, our final models accounted for SEER region. Concerning disease burden, adjustments for comorbidity, as well as tumor size and cancer stage when appropriate, were incorporated into the overall results.
There are other possible explanations why the vulnerable populations described here may receive inadequate care. At the level of the patient–physician relationship, either partner may influence what care is received. Among physicians or institutions, there may be bias in provision of clinical services based upon a patient's racial identity,34 relative lack of affluence, or even rural background. Access to clinical resources or less competence among physicians who serve vulnerable populations may also explain gaps in quality of care.35 Conversely, different patients may have different preferences for care, and the perceptions and decisions of individuals may be influenced by cultural beliefs.36 Family support mechanisms may further contribute to differences in care among vulnerable populations. Minority women, as well as women who live in rural areas and areas of lower socioeconomic status, may also seek care at more poorly performing medical institutions within a given region.
The process of documenting ER status has unique properties when considering explanations for any measured difference between groups. If the decision for surgery has been made, then assessment of ER status should routinely follow. It seems unlikely that patient preferences play any role in this regular process of care. Furthermore, if a patient has reached the point of surgery, access to the healthcare system (with or without insurance) no longer serves as a viable explanation for any gap in quality. Thus, the possible set of explanations for racial, ethnic, and rural differences in adequate documentation of ER status can be narrowed to either bias in the provision of services or gaps in quality of care at medical institutions predominantly serving these patients. Of course, this study measured adequate documentation of ER status, not necessarily performance of the test, but nonetheless, systematic differences in documentation of ER status reflect differential record-keeping at a minimum and, quite possibly, differential care.
The primary limitation of our study design to answer the questions posed here, is its potential for misclassification of different processes of care. SEER-Medicare data has been found to be reasonably accurate for determination of surgery,17, 18 XRT,20 and surveillance mammography.22 SEER is commonly the gold standard for ascertainment of ER status in population-based studies,37 and our definition of adequate documentation of ER status has face validity. Misclassification of race or ethnicity should also be considered. SEER data has previously been found to be accurate for the identification of Whites and African-Americans, and Hispanic and Asian/Pacific Islander ethnicity variables likely provide acceptable specificity.15 Another limitation is that there may have been insufficient power to detect differences between groups when no difference was found, particularly for analyses limited to breast cancers diagnosed from 1997 to 1999. Therefore, the relative improvement in the receipt of all processes of care among Hispanic women may have been due to insufficient sample sizes. Nonetheless, there was sufficient power to detect multiple differences in breast cancer care in more recent years, and in all instances where a difference was present, the pattern was consistently one of lower use among vulnerable populations. Finally, there are other processes of care besides the ones measured here that contribute to optimal breast cancer care, including axillary lymph lymph node dissection, chemotherapy, chemoprevention (tamoxifen), and other prognostic biomarkers (progesterone receptor and Her-2). If there is underuse of any of these care processes, then the findings described here may underestimate gaps in the quality of breast cancer care.
These results highlight the importance of a cancer data system to track healthcare quality and disparities across as broad a spectrum of the continuum of cancer care as possible. Another report by the Institute of Medicine suggested that current cancer data systems like SEER-Medicare while important are not ideal,38 and we would agree. In our view, at least two characteristics should be considered in the future when designing the ideal system. First, the ideal cancer data system should not only collect data that makes comparisons among vulnerable groups possible, but the data system should also routinely collect information that allows testing and monitoring suspected causes of these differences. Second, an ideal cancer data system would rapidly identify gaps in quality so that their recognition would make an immediate difference in the care of patients involved. Our findings of differences in breast cancer care among vulnerable populations demonstrate patterns largely consistent with previous literature. Importantly, this study provides information indicating that the problem of gaps in the quality of care among vulnerable populations, despite being identified as a target for intervention previously, persist and require bold, decisive action. Surveillance needs to be accompanied by interventions targeted both at minorities39 and other vulnerable populations, and evidence-based interventions should be disseminated widely.
- 10HewittM, SimoneJ, editors. Ensuring quality cancer care. Washington, DC: National Academy Press, 1999.
- 11Treatment of early stage breast cancer. NIH Consensus Statement Online 1990 Jun 18–21 [online] 8:1–19. Available at URL: http://consensus.nih.gov/cons/081/081_statement.htm[accessed February 10, 2005].
- 13National healthcare disparities report: summary. February 2004. Agency for Healthcare Research and Quality, Rockville, MD. Available at URL: http://www.ahrq.gov/qual/nhdr03/nhdrsum03.htm[accessed February 10, 2005].
- 14Overview of the SEER-Medicare data: content, research applications, and generalizability to the United States elderly population. Med Care. 2002; 40(8 Suppl): 3–18., , , et al.
- 15Patient demographic and socioeconomic characteristics in the SEER-Medicare database. Med Care. 2002; 40(8 Suppl): 19–25., , , et al.
- 17Use of SEER-Medicare data for measuring cancer surgery. Med Care. 2002; 40(8 Suppl): 43–47., , , et al.
- 20Studying radiation therapy using SEER-Medicare-linked data. Med Care. 2002; 40(8 Suppl): 49–54., , , et al.
- 22Methodological issues in the use of administrative claims data to study surveillance after cancer treatment. Med Care. 2002; 40(8 Suppl): 69–74., , , et al.
- 34SmedleyB, StithA, NelsonAR (eds). Unequal treatment: confronting racial and ethnic disparities in health care. Washington, DC: National Academy Press, 2003.
- 38HewittM, SimoneJ, editors. Enhancing data systems to improve the quality of cancer care. Washington, DC: National Academy Press, 2000.
- 39http://www.ahrq.gov/clinic/epcsums/minqusum.htm [accessed May 27, 2005]., , , et al. Strategies for Improving Minority Healthcare Quality. January 2004. Agency for Healthcare Research and Quality, Rockville, MD. Available at URL: