Abstract
This study examined racial and geographic disparities in breast cancer detection modalities (screening-, diagnostic-, or non-mammography) with cancer stage and mortality. A retrospective cohort study was conducted using Texas Cancer Registry-Medicare linkage data for geriatric women. Cancers detected through screening and diagnostic mammography had 43 % (95 % CI, 39 %–46 %, p < .0001) and 31 % (95 % CI, 27 %–35 %, p < .0001) lower all-cause mortality, and 49 % (95 % CI, 41 %–54 %, p < .0001) and 37 % (95 % CI, 32 %–43 %, p < .0001) lower cancer-specific mortality, respectively, compared to non-mammography-detected breast cancers. Patients from rural areas were 17 % (95 % CI, 1.06 – 1.29) more likely to be diagnosed with mid-(p = .0023) and advanced stage (p = .003) cancers compared to their urban counterparts. Racial or geographic disparities in cancer detection modalities with associated mortality no longer exist after adjusting for covariates. Healthcare professionals can leverage these findings to promote rural cancer health equity.
Keywords: Diagnostic mammography, Screening mammography, Breast cancer mortality, Racial disparity, Rural-urban disparity
Introduction
Breast cancer is the leading cause of non-dermatological cancer among women in the United States, with a yearly incidence rate of 127 per 100,000, and the second leading cause of cancer death in women.1 In 2023, updated USPSTF guidelines have recommended biennial mammography screening for women aged 40–74 years.2 Mammography is performed either for screening or diagnostic purpose. Screening mammography is used to check for breast cancer prior to development of any signs or symptoms of the disease with the goal of detecting cancer earlier.3 The benefits of screening mammography in breast cancer early detection and mortality reduction have been well-established.4,5 Diagnostic mammography is carried out typically following clinical signs and symptoms or as a follow up to an abnormal screening mammography.3,6,7 According to the Breast Imaging Reporting and Data System (BI-RADS), which was developed by the American College of Radiology (ACR) in the 1980s to improve the quality of mammography, screening and diagnostic mammography have different reporting systems.7 Additionally, the national performance benchmarks for screening and diagnostic mammography have been reported separately by the Breast Cancer Surveillance Consortium.6,8 However, there are limited studies that separately examine screening and diagnostic mammography utilization in breast cancer detection.
Racial and geographic disparities in mammography performance as well as breast cancer mortality have been reported. Recent studies found that rural Hispanic and non-Hispanic black (NHB) women in Texas had the lowest mammography screening rates among all racial groups.9,10 Urban women have higher screening rates and a lower chance of being diagnosed with late-stage cancer.11 Cancer-specific mortality and all-cause mortality rates for breast cancer have been found to be higher in rural women as compared to their urban counterparts.12–16 However, both screening and diagnostic mammography were either grouped together or undefined in these studies. It is unclear whether there are any racial/ethnic and geographic disparities in screening and diagnostic mammography utilization with associated cancer outcomes.
The purpose of this study was to examine racial/ethnic and geographic disparities in cancer detection modalities (screening, diagnostic, and non-mammography), along with their association with cancer stage at diagnosis, as well as all-cause and cancer-specific mortality, with and without adjusting for covariates (sociodemographic factors, comorbidities, and tumor characteristics).
Methods
Data source
We conducted a retrospective cross-sectional cohort study using Texas Cancer Registry (TCR)-Medicare linkage data from 2010–2019 with female aged 67 years and older who were diagnosed with breast cancers between 2010 and 2014. TCR is one of the largest cancer registries in the U.S. The data collected by TCR include patient demographics, primary tumor site, stage, first course of treatment, tumor morphology, cause of death, and survival status. In 2021, the TCR joined the Surveillance, Epidemiology and End Results (SEER) program. The TCR-Medicare linkage was performed by the National Cancer Institute and the Centers for Medicare & Medicaid Services (CMS). The Medicare claims data include billing information on inpatient services, physician services, and outpatient visits. For this study, data were extracted from Medicare beneficiary summary files, Medicare Provider Analysis and Review files, Outpatient Standard Analytical Files, and Medicare Carrier files. The study protocol has obtained the Institutional Review Board (IRB:22–0110) approval at the University of Texas Medical Branch.
Study cohort
To assess 5-year survival rate, we identified all female patients who were diagnosed with breast cancer for the first time at any point during the period of 2010–2014. We excluded females under 67 years old at the time of diagnosis, those with no continuous enrollment in Medicare Parts A & B, or those enrolled in a Health Maintenance Organization (HMO) within 2 years prior to breast cancer diagnosis. Additionally, we excluded beneficiaries categorized as ‘Non-Hispanic Others’ in terms of race/ethnicity due to a small sample size. Supplemental Figure 1 provides a flow chart of the study sample selection and the final cohort.
Study measures and covariates
The measures of interest were breast cancer detection modalities, cancer stage at diagnosis, and mortality in relation to race/ethnicity and geographic location. To assess breast cancer detection modalities, we examined the utilization of mammography within 2 years prior to breast cancer diagnosis. Breast cancers were categorized into three groups based on cancer detection modalities: diagnostic mammography detected, screening mammography detected, and non-mammography detected indicating that breast cancers were not detected through either diagnostic or screening mammography. As some patients underwent multiple mammograms (either diagnostic or screening) within 2 years prior to cancer diagnosis, Fenton’s algorithm was used to differentiate between diagnostic and screening purposes of the mammography.17,18 In essence, if an individual had two screening mammography within a 9-month period, the second mammogram was considered for diagnostic purpose. The underlying rationale is that the time interval for yearly mammography screening is usually greater than 9 months. For individuals who had diagnostic mammography only within 2 years prior to breast cancer diagnosis, the cancer was classified as diagnostic mammography detected. However, if an individual had only one screening mammogram, two screening mammograms with an interval greater than 9 months or one screening mammogram followed by a diagnostic mammogram, the mammogram was classified as being for screening purposes, and any detected cancer was considered as a result of screening mammography detected. Supplemental figure 2 illustrates the flowchart outlining the classification of breast cancer based on screening and diagnostic mammography utilization. Screening mammography was identified using the following Healthcare Common Procedure Code System (HCPCS) codes related to screening mammography performed between 2007 and 2014: 77,052, 77,057 and G0202. Diagnostic mammography was identified using the following HCPCS codes related to diagnostic mammography performed between 2007 and 2014: G0204 and G0206.
Breast cancers were staged using the Tumor, Node, Metastasis (TNM) staging system and further categorized into three groups: ‘Early stage’ including of “In-Situ” and “Localized” cancers; ‘Mid stage’ including of “Regional” cancers; and ‘Advanced stage’ including of “Distant- metastatic” and “Unstaged” cancers. The ‘Unstaged’ stage was categorized as advanced stage because the rates of unstaged cancer increase with age and race/ethnicity, and poor survival has been linked to unstaged diagnoses.19,20 Mortality outcomes included all-cause mortality and cancer-specific mortality within 5 years of diagnosis. Cancer mortality was identified using ICD-9 SEER specific cause of death codes for cancer (140–239 & 611) and ICD-10 SEER specific cause of death codes for cancer (C00-D489 & N61–64), with deaths from other causes censored.21
For all patients, we extracted sociodemographic factors including age, race-ethnicity, dual eligibility at diagnosis, marital status (defined as currently married at diagnosis), geographic areas (urban and rural areas), Medicare original entitlement (defined as disabled at enrollment), education and poverty at the zip code level. Race/Ethnicity was grouped into four categories: non-Hispanic whites, non-Hispanic blacks, Hispanics, and non-Hispanic others (Supplemental Figure 1). Non-Hispanic others were excluded from the study due to the small sample size. To distinguish geographic areas, we utilized the USDA’s 2013 Rural-Urban Continuum Codes (UUCC). Metropolitan counties were categorized based on the population size of their metro areas, while nonmetropolitan counties were classified according to their level of urbanization and proximity to a metro area.22 These codes were divided into two main categories: urban (UUCC codes 1–3) and rural (UUCC codes 4–9).
We also obtained Charlson’s comorbidity index (CCI) scores within 12 months prior to diagnosis, year of diagnosis, and cancer grades.23 Cancer grades were categorized into four groups based on cell differentiation: well differentiated, moderately differentiated, poorly/undifferentiated, and unknown (either missing or of unknown grade). Finally, we identified whether patients had an established oncology visit within the first year after diagnosis. An oncology visit was defined as an encounter with an oncologist who had CMS provider specialty codes 83, 90, 91, or 92. An established oncology service was defined as if a patient visited an oncologist on two or more occasions in an outpatient setting in the year after diagnosis based on evaluation and management (E&M) codes 99,201–99,205 (new patient encounters) and 99,211–99,215 (established patient encounters).24
Statistical analyses
We conducted comparisons of patients’ characteristics based on mammography utilization in breast cancer detection modalities, using Chi-square tests for categorical variables and ANOVA for continuous variables. Multinomial logistic regression analyses were used to examine racial/ethnic and geographic disparities in mammography utilization and cancer stage with and without adjusting for sociodemographic factors and comorbidity. The reference group was non-mammography-detected cancers and the one for stage outcome was ‘Early stage’. For the analysis of mammography utilization as an outcome of interest, the initial unadjusted multinomial regression model included race/ethnicity and geographic location (model I). Model I was then adjusted for sociodemographic factors (age, year of diagnosis, marital status, Medicare original reason at entitlement, dual eligibility, education level, poverty level) and CCI (model II). Regarding stage as an outcome of interest, the first unadjusted multinomial logistic regression model included race/ethnicity and geographic location (model I); model II comprised model I adjusted for sociodemographic factors and CCI.
All-cause mortality and cancer-specific mortality were estimated using failure curves derived from the Kaplan-Meier method. Subsequently, several Cox Proportional Hazard (pH) regression analyses with various adjustment were conducted to examine the associations between mortality and race/ethnicity or geographic location with and without adjusting for covariates of interest. For both all-cause and cancer-specific mortality, the first unadjusted Cox pH regression model I included race/ethnicity and geographic location (model I); model II comprised model I adjusted for cancer stage and grade; model III comprised model II further adjusted for sociodemographic factors, CCI, and oncology service. Patients were censored for a period of 5 years of post-diagnosis. The Proportional Hazard assumption for all the Cox pH regression models was visually assessed through a graph of the transform of the martingale residuals against follow-up time and confirmed using the Kolmogorov Supremum test.25 Collinearity among covariates was assessed using the variance inflation factor (VIF), and no serious multicollinearity was detected. All tests of statistical significance were 2-sided, and analyses were performed with SAS 9.4 (SAS Inc., Cary, NC).
Results
Descriptive characteristics of study population
A total of 21,293 female aged ≥ 67 were diagnosed with breast cancer between 2010–2014. The mean age of study sample was 76.9 (SD 7.04). Most patients were NHW (80 %), had moderately differentiated grade (33.3 %) of breast cancer, did not meet criteria for dual eligibility at enrollment (87.2 %), unmarried status (72.4 %), lived in urban areas (80.8 %), were not disabled (94 %), had an oncology visit within the first year of breast cancer diagnosis (83 %), and had a CCI score of zero (37 %) (Table 1A).
Table 1.
Characteristics ofbreast cancer patients (A) and outcomes (B) based on mammography utilization in breast cancer detection modalities among female (≥ 67 years) who were diagnosed with breast cancers between 2010–2014, TCR-Medicare linkage data from 2010–2019.
| A | |||||
|---|---|---|---|---|---|
|
| |||||
| Characteristics | Mammography Utilization |
||||
| Total N = 21,293 N ( %) |
None 6498 (30.5 % )N ( %) |
Diagnostic 7424 (34.9 %) N ( %) |
Screening 7371 (34.6 %) N ( %) |
P | |
|
| |||||
| Age | .0001 | ||||
| Mean, (SD) | 76.9 (7.04) | 78.9 (7.98) | 76.5 (6.64) | 75.5 (6.07) | |
| Race/Ethnicity | .0001 | ||||
| Hispanic | 2571 (12.1) | 925 (36) | 861 (33.5) | 785 (30.5) | |
| NHW | 17,041 (80) | 5012 (29.4) | 5993 (35.2) | 6036 (35.4) | |
| NHB | 1681 (7.9) | 561 (33.4) | 570 (33.9) | 550 (32.7) | |
| Year ofDiagnosis | .4002 | ||||
| 2010 | 4506 (21.2) | 1383 (30.7) | 1590 (35.3) | 1533 (34) | |
| 2011 | 4372 (20.5) | 1335 (30.5) | 1497 (34.2) | 1540 (35.2) | |
| 2012 | 4433 (20.8) | 1310 (29.6) | 1567 (35.3) | 1556 (35.1) | |
| 2013 | 3936 (18.5) | 1177 (29.9) | 1390 (35.3) | 1369 (34.8) | |
| 2014 | 4046 (19) | 1293 (32) | 1380 (34.1) | 1373 (33.9) | |
| Cancer Grade | .0001 | ||||
| Well differentiated | 3737 (17.5) | 817 (21.9) | 1447 (38.7) | 1473 (39.4) | |
| Moderately differentiated | 7088 (33.3) | 2030 (28.6) | 2464 (34.8) | 2594 (36.6) | |
| Poorly/Undifferentiated | 4907 (23.1) | 1553 (31.6) | 1663 (33.9) | 1691 (34.5) | |
| Unknown | 5561 (26.1) | 2098 (37.7) | 1850 (33.3) | 1613 (29) | |
| Dual Eligibility | .0001 | ||||
| Yes | 2724 (12.8) | 1186 (43.5) | 844 (31) | 694 (25.5) | |
| No | 18,569 (87.2) | 5312 (28.6) | 6580 (35.4) | 6677 (36) | |
| Married | .0001 | ||||
| Yes | 5868 (27.6) | 1399 (23.8) | 2076 (35.4) | 2393 (40.8) | |
| No | 15,425 (72.4) | 5099 (33) | 5348 (34.7) | 4978 (32.3) | |
| Geographic areas | .0053 | ||||
| Urban | 17,215 (80.8) | 5168 (30) | 6048 (35.1) | 5999 (34.9) | |
| Rural | 4078 (19.2) | 1330 (32.6) | 1376 (33.7) | 1372 (33.6) | |
| Disabled at Entitlement | .0001 | ||||
| Yes | 1270 (6) | 452 (35.6) | 455 (35.8) | 363 (28.6) | |
| No | 20,023 (94) | 6046 (30.2) | 6969 (34.8) | 7008 (35) | |
| Quartiles of % with < 12 years of education | .0001 | ||||
| 0–12 (Very High) | 5104 (25.1) | 1328 (26) | 1838 (36) | 1938 (38) | |
| 12–21 (High) | 5084 (24.9) | 1420 (27.9) | 1848 (36.4) | 1816 (35.7) | |
| 21 –30 (Moderate) | 5101 (25.1) | 1643 (32.3) | 1769 (34.6) | 1689 (33.1) | |
| 30–100 (Poor) | 5078 (24.9) | 1824 (35.9) | 1645 (32.4) | 1609 (31.7) | |
| Quartile of % residents living below poverty | .0001 | ||||
| 0–8 (Very low) | 5124 (25.1) | 1357 (26.5) | 1864 (36.4) | 1903 (37.1) | |
| 8–12 (Low) | 5065 (24.9) | 1459 (28.8) | 1762 (34.8) | 1844 (36.4) | |
| 12–19 (Moderate) | 5097 (25) | 1609 (31.6) | 1757 (34.5) | 1731 (33.9) | |
| 19–100 (High) | 5081 (25) | 1790 (35.2) | 1717 (33.8) | 1574 (31) | |
| Oncology Service | .0001 | ||||
| Yes | 17,675 (83) | 4648 (26.3) | 6447 (36.5) | 6580 (37.2) | |
| No | 3618 (17) | 1850 (51.1) | 977 (27.0) | 791 (21.9) | |
| CCI | .0001 | ||||
| 0 | 7880 (37) | 2310 (29.3) | 2653 (33.7) | 2917 (37.0) | |
| 1 | 6022 (28.3) | 1678 (27.9) | 2149 (35.7) | 2195 (36.4) | |
| ≥2 | 7391 (34.7) | 2510 (33.9) | 2622 (35.5) | 2259 (30.6) | |
|
| |||||
| B | |||||
|
| |||||
| Outcomes | Mammography Utilization |
||||
| Total N = 21,293 N (%) |
None N = 6498 N (%) |
Diagnostic N = 7424 N (%) |
Screening N = 7371 N ( %) |
P | |
|
| |||||
| Cancer Stage | .0001 | ||||
| Early stage | 13,321 (62.6) | 2935 (45.2) | 5069 (68.3) | 5317 (72.1) | |
| Mid stage | 4044 (19) | 1601 (24.6) | 1241 (16.7) | 1202 (16.3) | |
| Advanced stage | 3928 (18.4) | 1962 (30.2) | 1114 (15) | 852 (11.6) | |
| 5-year Mortality | |||||
| All-cause | 6964 (32.7) | 3320 (51.1) | 2066 (27.8) | 1578 (21.4) | .0001 |
| Cancer-specific | 3421 (16.1) | 1883 (29) | 893 (12) | 645 (8.8) | .0001 |
Abbreviations: SD, Standard Deviation; BC, Breast Cancer; NHW, non-Hispanic white; NHB, non-Hispanic black; CCI, Charlson Comorbidity Index.
The characteristics of the study population were also summarized based on breast cancer detection modality (Table 1A). Among the study population, 34.6 % were detected through screening mammography, 34.9 % through diagnostic mammography, and 30.5 % through modalities other than mammography. Among Hispanic women, the majority of breast cancers (36 %) were non-mammography detected, compared to 33.4 % among NHB women and 29.4 % among NHW women. Additionally, Hispanic women showed the lowest percentage of breast cancers detected through screening mammography (30.5 %), compared to 32.7 % among NHB women and 35.4 % among NHW women. Overall, breast cancer patients diagnosed through non-mammography tended to be older, have higher-grade tumors, and show a higher percentage of individuals with dual eligibility at enrollment, unmarried status, disability, lower education levels, increased poverty, lack of oncologist services, and higher CCI scores compared to patients diagnosed through screening or diagnostic mammography (Table 1A).
Next, we summarized the characteristics of breast cancer stage at diagnosis and 5-year all-cause and cancer-specific mortality based on breast cancer detection modality (Table 1B). Overall, 62.6 % of breast cancers were diagnosed at an early stage. Specifically, 45.2 % non-mammography detected, 68.3 % diagnostic mammography detected, and 72.1 % screening mammography detected breast cancers were at early-stage. 30.2 % of non-mammography detected breast cancers were at advanced-stage, compared to 15 % diagnostic mammography detected, and 11.6 % screening mammography-detected breast cancers. The 5-year all-cause mortality rates were 51.1 % for non-mammography detected breast cancers, compared to 27.8 % and 21.4 % for diagnostic and screening mammography detected breast cancers (p < .0001), respectively. Similarly, the 5-year cancer-specific mortality rates were 29 % for non-mammography detected breast cancers, compared to 12 % and 8.8 % for diagnostic and screening mammography detected breast cancers (p < .0001), respectively.
Racial/ethnic disparities in cancer detection modality, cancer stage at diagnosis, and mortality
Compared to NHW, NHB breast cancers were 16 % (95 % CI, 5 %–26 %, p = .0047) and 20 % (95 % CI, 8 %–29 %, p = .0005) less likely to be detected by diagnostic and screening mammography, respectively (Table 2 Model I). Additionally, NHB breast cancers were 21 % (95 % CI, 1.06–1.38, p = .0042) and 36 % (95 % CI, 1.19–1.55, p < .0001) more likely to be diagnosed at mid-stage and advanced stage, respectively (Table 3 Model I). NHB patients also had 38 % (95 % CI, 1.27–1.49, p < .0001) and 39 % (95 % CI, 1.24–1.55, p < .0001) higher all-cause and cancer-specific mortality, respectively, compared to NHW (Table 4 Model I). Similarly, compared to NHW, Hispanics breast cancers were 23 % (95 % CI, 15 %–30 %, p < .0001) and 30 % (95 % CI, 22 %–37 %, p < .0001) less likely to be detected by diagnostic and screening mammography (Table 2 Model I). They were also 25 % (95 % CI of OR, 1.12–1.39, p < .0001) and 13 % (95 % CI of OR, 1.01–1.26, p = .037) more likely to be diagnosed at mid-stage and advanced stage, respectively (Table 3 Model I). Hispanic patients had 9 % (95 % CI of HR, 1.01–1.16, p = .0237) higher all-cause mortality but no significance difference in cancer-specific mortality (Table 4 Model I).
Table 2.
Multinominal logistic regression assessing racial/ethnic and geographic disparities in mammography utilization for breast cancer diagnosis.
| Mammography Utilization (Ref=None) | Model I |
Model II |
||||||
|---|---|---|---|---|---|---|---|---|
| OR | 95 % CI | p | OR | 95 % CI | p | |||
|
| ||||||||
| Race/Ethnicity (Ref=NHW) | ||||||||
| Hispanic/Diagnostic | 0.77 | 0.70 | 0.85 | <0.0001 | 0.94 | 0.83 | 1.06 | .2953 |
| Hispanic/Screening | 0.70 | 0.63 | 0.78 | <0.0001 | 0.90 | 0.80 | 1.02 | .0963 |
| NHB/Diagnostic | 0.84 | 0.74 | 0.95 | .0047 | 1.02 | 0.89 | 1.17 | .8194 |
| NHB/Screening | 0.80 | 0.71 | 0.91 | .0005 | 1.08 | 0.94 | 1.25 | .2706 |
| Geographic areas (Ref=Urban) | ||||||||
| Rural/Diagnostic | 0.87 | 0.80 | 0.94 | .001 | 0.97 | 0.88 | 1.07 | .5381 |
| Rural/Screening | 0.87 | 0.80 | 0.94 | .001 | 0.98 | 0.89 | 1.08 | .739 |
Abbreviations: OR, odds ratio; CI, confidence interval; NHW, non-Hispanic white; NHB, non-Hispanic black.
Model 1: Unadjusted analysis including race/ethnicity and geographic areas.
Model 2: Model 1 adjusted for age, year of breast diagnosis, marital status, original reason at entitlement, dual eligibility, education level, poverty level and Charlson Comorbidity Index.
Table 3.
Multinominal logistic regression assessing racial/ethnic and geographic disparities in breast cancer stage at diagnosis with and without adjusting for covariates.
| Stage category (Ref=Early stage) | OR | 95 % CI | p | |
|---|---|---|---|---|
| Model I | ||||
|
| ||||
| Race/Ethnicity (Ref=NHW) | ||||
| Hispanic/Mid stage | 1.25 | 1.12 | 1.39 | <0.0001 |
| Hispanic/Advanced stage | 1.13 | 1.01 | 1.26 | .037 |
| NHB/Mid stage | 1.21 | 1.06 | 1.38 | .0042 |
| NHB/Advanced stage | 1.36 | 1.19 | 1.55 | <0.0001 |
| Geographic areas (Ref=Urban) | ||||
| Rural/Mid stage | 1.23 | 1.12 | 1.34 | <0.0001 |
| Rural/Advanced stage | 1.28 | 1.17 | 1.40 | <0.0001 |
| Model II | ||||
| Race/Ethnicity (Ref=NHW) | ||||
| Hispanic/Mid stage | 1.06 | 0.93 | 1.20 | .3706 |
| Hispanic/Advanced stage | 0.95 | 0.83 | 1.08 | .4234 |
| NHB/Mid stage | 1.00 | 0.87 | 1.16 | .9545 |
| NHB/Advanced stage | 1.05 | 0.90 | 1.21 | .5464 |
| Geographic areas (Ref=Urban) | ||||
| Rural/Mid stage | 1.17 | 1.06 | 1.29 | .0023 |
| Rural/Advanced stage | 1.17 | 1.05 | 1.29 | .003 |
Abbreviations: OR, odds ratio; CI, confidence interval; NHW, non-Hispanic white; NHB, non-Hispanic black; Ref, reference.
Model I: Unadjusted analysis including race-ethnicity and urban-rural status.
Model II: Model I, adjusted for age, year of BC diagnosis, marital status, original reason at entitlement, dual eligibility, education level, poverty level, CCI.
Table 4.
Cox proportional hazards regression analyses assessing the racial/ethnic and geographic disparities in 5-year all-cause and cancer-specific mortality.
| All-cause mortality |
Cancer-specific mortality |
|||||||
|---|---|---|---|---|---|---|---|---|
| HR | 95 %CI | p | HR | 95 % CI | p | |||
| Model I | Model I | |||||||
|
| ||||||||
| Race/Ethnicity (Ref=NHW) | ||||||||
| Hispanic | 1.09 | 1.01 | 1.16 | .0237 | 1.08 | 0.98 | 1.20 | .1272 |
| NHB | 1.38 | 1.27 | 1.49 | <0.0001 | 1.39 | 1.24 | 1.55 | <0.0001 |
| Geographic areas (Ref=Urban) | ||||||||
| Rural | 1.12 | 1.06 | 1.19 | <0.0001 | 1.18 | 1.09 | 1.28 | <0.0001 |
| Model II | Model II | |||||||
| Race/Ethnicity (Ref=NHW) | ||||||||
| Hispanic | 1.02 | 0.95 | 1.10 | .5739 | 0.99 | 0.89 | 1.09 | .8149 |
| NHB | 1.24 | 1.15 | 1.35 | <0.0001 | 1.17 | 1.05 | 1.31 | .0053 |
| Geographic areas (Ref=Urban) | ||||||||
| Rural | 1.05 | 0.99 | 1.11 | .1209 | 1.07 | 0.99 | 1.16 | .1113 |
| Model III | Model III | |||||||
| Race/Ethnicity (Ref=NHW) | ||||||||
| Hispanic | 0.90 | 0.83 | 0.98 | .0114 | 0.92 | 0.82 | 1.04 | .1920 |
| NHB | 1.01 | 0.93 | 1.11 | .7785 | 1.07 | 0.94 | 1.21 | .3039 |
| Geographic areas (Ref=Urban) | ||||||||
| Rural | 0.93 | 0.87 | 0.99 | .0267 | 0.98 | 0.89 | 1.07 | .5774 |
Abbreviations: HR, hazard ratio; Cl, confidence interval; NHW, non-Hispanic white; NHB, non-Hispanic black.
Model I: Unadjusted analysis including race-ethnicity and urban-rural status.
Model II: Model I adjusted for cancer stage and grade.
Model III: Model II adjusted for age, year of breast cancer diagnosis, marital status, original reason at entitlement, dual eligibility, education level, poverty level, CCI, oncologist visit within first year of diagnosis.
After adjusting for cancer stage and grade, NHB cancer patients still had 24 % (95 % CI of HR, 1.15–1.35, p < .0001) and 17 % (95 % CI of HR, 1.05–1.31, p < .0001) higher all-cause and cancer-specific mortality, respectively, compared to NHW (Table 4 Model II). There were no significant disparities in breast cancer detection modality, cancer stage at diagnosis, all-cause mortality, or cancer-specific mortality across all race/ethnicity groups after adjusting for covariates (Table 2 Model II, Table 3 Model II, and Table 4 Model III). Interestingly, Hispanic breast cancer patients had 10 % (95 % CI, 2 %–17 %, p = .0114) lower all-cause mortality compared to NHW after adjusting for covariates (Table 4 Model III).
Geographic disparities in cancer detection modality, cancer stage at diagnosis, and mortality
Rural women’s breast cancers were 13 % less likely to be detected either by diagnostic or by screening mammography (95 % CI, 6 %–20 % p = .001) (Table 2 Model I), were 23 % (95 % CI of OR, 1.12–1.34, p < .0001) and 28 % (95 % CI of OR, 1.17–1.4, p < .0001) more likely to be diagnosed at mid-stage and advanced-stage (Table 3 Model I), and had 12 % (95 % CI of HR, 1.06–1.19, p < .0001) and 18 % (95 % CI of HR, 1.09–1.28, p < .0001) higher all-cause and cancer-specific mortality (Table 4 Model I), respectively, compared to NHW.
After adjusting for covariates, patients from rural areas were 17 % more likely to be diagnosed with mid-stage (95 % CI of OR, 1.06 – 1.29, p = .0023) and advanced-stage (95 % CI of OR, 1.05–1.29, p = .003) (Table 3 Model II), and had 7 % (95 % CI, 1 %–13 %, p = .0184) lower all-cause mortality compared to their counterparts in urban areas (Table 4 Model III). No geographic disparities were observed in breast cancer detection modality and cancer-specific mortality after adjusting for covariates (Table 2 Model II & Table 4 Model III).
Discussion
In the present study, approximately two-thirds of diagnostic mammography detected and three-quarters of screening mammography detected breast cancers were at early stage, which is consistent with the benchmarks reported by the Breast Cancer Surveillance Consortium in 2017 (63.4 % for diagnostic and 76.9 % for screening mammography).6,6 Overall, approximately one-third of breast cancers were screening mammography detected (asymptomatic) and another third through diagnostic mammography (symptomatic). The non-mammography detected breast cancers could include both symptomatic and asymptomatic cancers (e.g., incidental detection by imaging unrelated to breast symptomatology). Breast cancers detected through diagnostic mammography may exhibit clinical similarities to symptomatic interval cancers. For instance, Kramp et al. reported an increased presence of prognostically unfavorable molecular subtypes in cases detected through diagnostic mammography.26 Additionally, Koo et al. reported a significant variation among patients diagnosed with advanced stage disease (e.g., stage IV) based on presenting symptoms.27 They also found common presenting breast cancer symptoms (e.g., breast lump) were linked to reduced likelihoods of advanced disease.27 These findings highlight opportunities for interventions to improve early detection of breast cancer through screening and to identify symptomatic breast cancer through diagnostic mammography.
No significant racial/ethnic differences were found in the modality (diagnostic-, screening-, or non-mammography) of breast cancer detection, cancer stages, and mortality rates after adjusting for covariates. In contrast to our study, Nyante et al. reported that NHB women were more likely to have cancer detected through diagnostic mammography after adjusting for breast cancer risk factors, mammogram modality, demographics, imaging, and imaging facility.28 Another study revealed that diagnostic delays following an abnormal screening mammogram were significantly associated with NHB or Hispanics, even after controlling for age, breast cancer history, and marital status.29 It’s worth noting that the covariates in those studies differed from those in this study, which might contribute to the variations in the findings. The findings of this study suggest that sociodemographic factors, tumor characteristics, and comorbidities play a significant role in the existing disparities in breast cancer measures including diagnosis modality, cancer stage at diagnosis, and mortality. Beyond race and ethnicity, other sociodemographic factors—such as low socioeconomic status, including low income, limited education, and language barriers—are well-established risk factors for disparities across the cancer care continuum.30 Additionally, NHB patients are more likely to be diagnosed with triple-negative breast cancer, an aggressive subtype that affects screening detectability.31 Data from Medicare beneficiaries in the U.S indicate that 40 % cancer patients have at least one other chronic condition.32 Breast cancer patients with comorbidities are less likely to receive curative treatment.33,34
After adjusting for covariates, we observed rural-urban disparities in cancer stages at diagnosis but found no disparities in mortality or cancer detection modality. To the best of our knowledge, this is the first study to separately examine rural-urban disparities in cancer detection by screening versus diagnostic mammography. Our findings on cancer stage at diagnosis and mortality align with previous research. For example, LeBlanc et al. recently reported that rural patients were more likely to be diagnosed with later-stage breast cancer, even after adjusting for age, race, tumor grade, receptor status, and insurance status.35 Similarly, Blake et al. found no significant rural-urban disparities in breast cancer mortality.36 Our findings suggest that the rural-urban disparities in mortality among breast cancer patients are likely attributed to sociodemographic factors, comorbidities, and tumor characteristics. It is well-recognized that rural populations, in general, tend to be older, more economically disadvantaged, and in poorer health compared to urban populations.37 Given the rural-urban differences across states, it’s also common to see higher cancer-specific and all-cause mortality rates for breast cancer among rural women when compared to their urban counterparts.12–16,38 The variations in these findings may not only be attributed to disparities in cancer diagnosis and treatments but also to differences in sociodemographic factors.
Awareness of disparities in breast cancer detection modality, cancer stage at diagnosis, and all-cause and cancer-specific mortality have clinical implications. Policymakers can more effectively allocate resources to rural areas and underserved populations ensuring equitable access to healthcare services. Strategies may include expanding mobile and community-based screening programs, broadening the reach of the National Breast and Cervical Cancer Early Detection Program (NBCCEDP), increasing Medicaid coverage and financial assistance for mammograms, and improving transportation access. Healthcare professionals can provide personalized medical care, considering racial/ethnic and geographic disparities, to promote high-quality care. Given the various barriers to breast cancer screening among rural and minority populations (e.g., longer travel distances to mammography centers, lower education levels, and lack of insurance),39,40 potential mitigation strategies should focus on addressing access barriers (e.g., transportation), as well as social determinants of health and system-level factors (e.g., insurance coverage). Medicaid expansion and other publicly funded health insurance programs for low-income individuals have been shown to increase cancer screening rates among underserved populations.41 Overall, awareness of these disparities leads to a more equitable healthcare system, ensuring all individuals have the opportunity for early breast cancer detection and improved survival outcomes.
Furthermore, a growing body of evidence recognizes potential negative consequences of mammography screening, such as overdiagnosis, false-positives, anxiety, and radiation exposure, particularly among older women.42,43 Therefore, there may be a need to reevaluate mammography screening practices to balance the benefits and harms of overuse among geriatric populations. Decisions regarding mammography recommendations should be made with careful consideration of the life expectancy and quality of life of older women.43,44
Some of the limitations should be acknowledged. The use of TCR-Medicare linkage data focusing on cancer patients in Texas restricts the generalizability of the findings to the national or international level. Additionally, a cross-sectional analysis of screening and diagnostic mammography utilization within two years before cancer diagnosis cannot fully capture an individual’s long-term mammography screening behavior or distinguish interval cancers, especially for those whose cancer was detected through diagnostic mammography. Furthermore, as this study is based on pre-pandemic data, the cancer outcomes do not reflect potential changes in mammography screening policies during the pandemic. Therefore, the impact of COVID-related policies on breast cancer outcomes could not be assessed. Future research should consider evaluating the long-term mammography screening behavior before cancer diagnosis and symptom presentation at the initial diagnosis of breast cancer. This approach could assist in better distinguishing interval cancers and assessing the correlation between diagnostic modalities and breast cancer outcomes.
Conclusions
This study demonstrates strong correlations between breast cancer detection modality (screening, diagnostic, and non-mammography) and cancer stage at diagnosis, as well as all-cause and cancer-specific mortality rates after adjusting for covariates. We observed significant geographic disparities in cancer stage at diagnosis, but not in mortality, after adjusting for covariates. These findings highlight the importance of addressing geographic impacts on breast cancer health disparities among geriatric populations.
Supplementary Material
Supplementary materials
Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.gerinurse.2025.103514.
Funding
This work was supported by Data Management and Analysis Core for Comparative Effectiveness Research on Cancer in Texas (RP 210130) funded by Cancer Prevention and Research Institute of Texas. S.G. reports funding from Komen SAC150061, CPRIT RP160674, CPRIT RP210140, and National Cancer Institute P30 CA016672. Z.L. reports funding from National Cancer Institute RO3CA292975.
Footnotes
Declaration of competing interest
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
CRediT authorship contribution statement
Zhaoli Liu: Conceptualization, Methodology, Writing – review & editing, Validation, Funding acquisition, Writing – original draft, Project administration. Mathilda Nicot-Cartsonis: Validation, Writing – original draft. Biai D.E. Digbeu: Writing – original draft, Data curation, Formal analysis, Methodology. Daoqi Gao: Formal analysis, Data curation, Methodology. Sharon H. Giordano: Funding acquisition, Methodology, Validation, Supervision, Conceptualization, Writing – review & editing. Yong-Fang Kuo: Data curation, Writing – review & editing, Methodology, Validation, Funding acquisition, Conceptualization, Supervision.
References
- 1.Centers For Disease Control and Prevention. United States Cancer Statistics. June 23, 2023. Accessed July 19, 2023. https://www.cdc.gov/cancer/uscs/index.htm.
- 2.United States Preventive Services Taskforce. Draft recommendation: breast cancer: screening. Accessed July 19, 2023. https://www.uspreventiveservicestaskforce.org/uspstf/draft-recommendation/breast-cancer-screening-adults.
- 3.National Cancer Institute. Mammograms. February 21, 2023. Accessed July 19, 2023. https://www.cancer.gov/types/breast/mammograms-fact-sheet.
- 4.Duffy SW, Tabár L, Yen AM, et al. Mammography screening reduces rates of advanced and fatal breast cancers: results in 549,091 women. Cancer. 2020;126(13):2971–2979. 10.1002/cncr.32859. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Duffy SW, Vulkan D, Cuckle H, et al. Effect of mammographic screening from age 40 years on breast cancer mortality (UK Age trial): final results of a randomised, controlled trial. Lancet Oncol. 2020;21(9):1165–1172. 10.1016/S1470-2045(20)30398-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Sprague BL, Arao RF, Miglioretti DL, et al. National Performance Benchmarks for modern diagnostic digital mammography: update from the Breast Cancer Surveillance Consortium. Radiology. 2017;283(1):59–69. 10.1148/radio-l.2017161519. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.American College of Radiology. Breast imaging reporting & data system. Accessed July 26, 2023. https://www.acr.org/Clinical-Resources/Reporting-and-Data-Systems/Bi-Rads.
- 8.Lehman CD, Arao RF, Sprague BL, et al. National Performance Benchmarks for modern screening digital mammography: update from the Breast Cancer Surveillance Consortium. Radiology. 2017;283(1):49–58. 10.1148/radio-l.2016161174. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Liu Z, Kuo YF, Giordano SH. Disparities in screening mammography utilization among Hispanic women in rural Texas from 2002 to 2018. Cancer Causes Control. Publ. online June 24. 2023. 10.1007/s10552-023-01738-3. [DOI] [Google Scholar]
- 10.Liu Z, Shan Y, Kuo YF, Giordano SH. Rural racial disparities and barriers in mammography utilization among Medicare beneficiaries in Texas: a longitudinal study. Geriatr Nur (L.). 2024;55:14–20. 10.1016/j.gerinurse.2023.10.019. [DOI] [Google Scholar]
- 11.Chandak A, Nayar P, Lin G. Rural-urban disparities in access to breast cancer screening: a spatial clustering analysis. J Rural Health. 2019;35(2):229–235. 10.1111/jrh.12308. [DOI] [PubMed] [Google Scholar]
- 12.Obeng-Gyasi S, Timsina L, Bhattacharyya O, Fisher CS, Haggstrom DA. Breast cancer presentation, surgical management and mortality across the rural–Urban continuum in the National cancer Database. Ann Surg Oncol. 2020;27(6):1805–1815. 10.1245/s10434-020-08376-y. [DOI] [PubMed] [Google Scholar]
- 13.Kenzik KM, Rocque GB, Landier W, Bhatia S. Urban versus rural residence and outcomes in older patients with breast cancer. Cancer Epidemiol Biomark. Prev. 2020;29(7):1313–1320. 10.1158/1055-9965.EPI-19-1414. [DOI] [Google Scholar]
- 14.Chu QD, Hsieh MC, Chu Y, et al. Do rural patients with operable breast cancer fare worse than urban patients in Louisiana? Results of the Louisiana cancer consortium. Surgery. 2020;168(4):653–661. 10.1016/j.surg.2020.04.053. [DOI] [PubMed] [Google Scholar]
- 15.Obeng-Gyasi S, Obeng-Gyasi B, Tarver W. Breast cancer disparities and the impact of geography. Surg Oncol Clin. 2022;31(1):81–90. 10.1016/j.soc.2021.08.002. [DOI] [Google Scholar]
- 16.Ho KL, Shiels MS, Ramin C, et al. County-level geographic disparities in cardiovascular disease mortality among US breast cancer survivors, 2000–2018. JNCI Cancer Spectr. 2023;7(1):pkac083. 10.1093/jncics/pkac083. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Fenton JJ, Zhu W, Balch S, Smith-Bindman R, Fishman P, Hubbard RA. Distinguishing screening from diagnostic mammograms using medicare claims data. Med Care. 2014;52(7):e44–e51. 10.1097/MLR.0b013e318269e0f5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Fenton JJ, Onega T, Zhu W, et al. Validation of a Medicare claims-based algorithm for identifying breast cancers detected at screening mammography. Med Care. 2016;54(3):e15–e22. 10.1097/MLR.0b013e3182a303d7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Merrill RM, Sloan A, Anderson AE, Ryker K. Unstaged cancer in the United States: a population-based study. BMC Cancer. 2011;11(1):402. 10.1186/1471-2407-11-402. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Flores R, Patel P, Alpert N, Pyenson B, Taioli E. Association of stage shift and population mortality among patients with non–Small cell lung cancer. JAMA Netw Open. 2021;4(12):e2137508. 10.1001/jamanetworkopen.2021.37508. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.ICD-9 SEER cause-specific death for sequence 00 - SEER recodes. Accessed May 25, 2023. https://seer.cancer.gov/causespecific/icd9seq00.html.
- 22.U.S. Department of Agriculture. USDA rural-urban continuum codes. Accessed March 25, 2023. https://www.ers.usda.gov/data-products/rural-urban-continuum-codes.aspx.
- 23.National Cancer Institute. SEER-Medicare Linked Data Resource. Accessed June 26, 2023. https://healthcaredelivery.cancer.gov/seermedicare/.
- 24.Shah BR, Hux JE, Laupacis A, Zinman B, Cauch-Dudek K, Booth GL. Administrative data algorithms can describe ambulatory physician utilization. Health Serv Res. 2007;42(4):1783–1796. 10.1111/j.1475-6773.2006.00681.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Copyright © SAS Institute Inc. All rights reserved. SAS Help Center: Assessment of the Proportional Hazards Model. 2019. Accessed January 29, 2023. https://documentation.sas.com/doc/en/pgmsascdc/9.4_3.4/statug/statug_phreg_details75.htm. [Google Scholar]
- 26.Kramp LJ, Mathiak M, Behrens HM, Schäfer FW, van Mackelenbergh M, Röcken C. The age-specific differences in histopathological tumor characteristics and TNM classification of breast carcinomas in Quality assured mamma diagnostic (Qua-MaDi) program in the state of Schleswig-Holstein in Germany. J Cancer Res Clin Oncol. 2022;148(2):387–400. 10.1007/s00432-021-03841-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Koo MM, Swann R, McPhail S, et al. Presenting symptoms of cancer and stage at diagnosis: evidence from a cross-sectional, population-based study. Lancet Oncol. 2020;21(1):73–79. 10.1016/S1470-2045(19)30595-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Nyante SJ, Abraham L, Aiello Bowles EJ, et al. Diagnostic mammography performance across racial and ethnic groups in a national network of community-based breast imaging facilities. Cancer Epidemiol Biomark. Prev. 2022;31(7):1324–1333. 10.1158/1055-9965.EPI-21-1379. [DOI] [Google Scholar]
- 29.Wujcik D, Shyr Y, Li M, et al. Delay in diagnostic testing after abnormal mammography in low-income women. Oncol Nurs Forum. 2009;36(6):709. 10.1188/09.ONF.709-715. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Singh S, Sridhar P. A narrative review of sociodemographic risk and disparities in screening, diagnosis, treatment, and outcomes of the most common extrathoracic malignancies in the United States. J Thorac Dis. 2021;13(6):3827–3843. 10.21037/jtd-21-87. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Moss JL, Tatalovich Z, Zhu L, Morgan C, Cronin KA. Triple-negative breast cancer incidence in the United States: ecological correlations with area-level sociodemographics, healthcare, and health behaviors. Breast Cancer. 2021;28(1):82–91. 10.1007/s12282-020-01132-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Edwards BK, Noone AM, Mariotto AB, et al. Annual Report to the Nation on the status of cancer, 1975–2010, featuring prevalence of comorbidity and impact on survival among persons with lung, colorectal, breast, or prostate cancer. Cancer. 2014;120(9):1290–1314. 10.1002/cncr.28509. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Land LH, Dalton SO, Jensen MB, Ewertz M. Influence of comorbidity on the effect of adjuvant treatment and age in patients with early-stage breast cancer. Br J Cancer. 2012;107(11):1901–1907. 10.1038/bjc.2012.472. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Sarfati D, Koczwara B, Jackson C. The impact of comorbidity on cancer and its treatment. CA Cancer J Clin. 2016;66(4):337–350. 10.3322/caac.21342. [DOI] [PubMed] [Google Scholar]
- 35.LeBlanc G, Lee I, Carretta H, Luo Y, Sinha D, Rust G. Rural-urban differences in breast cancer stage at diagnosis. Womens Health Rep New Rochelle N. 2022;3(1):207–214. 10.1089/whr.2021.0082. [DOI] [Google Scholar]
- 36.Blake KD, Moss JL, Gaysynsky A, Srinivasan S, Croyle RT. Making the case for investment in rural cancer control: an analysis of rural cancer incidence, mortality, and funding trends. Cancer Epidemiol Biomark Prev Publ Am Assoc Cancer Res Cosponsored Am Soc Prev Oncol. 2017;26(7):992–997. 10.1158/1055-9965.EPI-17-0092. [DOI] [Google Scholar]
- 37.American Hospital Association. Rural report: challenges facing Rural communities and the roadmap to ensure local access to high-quality, affordable care. Published online 2019. [Google Scholar]
- 38.Sprague BL, Ahern TP, Herschorn SD, Sowden M, Weaver DL, Wood ME. Identifying key barriers to effective breast cancer control in rural settings. Prev Med. 2021;152 (Pt 2):106741. 10.1016/j.ypmed.2021.106741. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Jewett PI, Gangnon RE, Elkin E, et al. Geographic access to mammography facilities and frequency of mammography screening. Ann Epidemiol. 2018;28(2):65–71.e2. 10.1016/j.annepidem.2017.11.012. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.United States Census Bureau. Health insurance in Rural America. Accessed March 28, 2025. https://www.census.gov/library/stories/2019/04/health-insurance-rural-america.html.
- 41.Hendryx M, Luo J. Increased cancer screening for low-income adults under the Affordable Care Act Medicaid expansion. Med Care. 2018;56(11):944–949. 10.1097/MLR.0000000000000984. [DOI] [PubMed] [Google Scholar]
- 42.Richman IB, Long JB, Soulos PR, Wang SY, Gross CP. Estimating breast cancer overdiagnosis after screening mammography among older women in the United States. Ann Intern Med. 2023;176(9):1172–1180. 10.7326/M23-0133. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Canelo-Aybar C, Posso M, Montero N, et al. Benefits and harms of annual, biennial, or triennial breast cancer mammography screening for women at average risk of breast cancer: a systematic review for the European Commission Initiative on Breast Cancer (ECIBC). Br J Cancer. 2022;126(4):673–688. 10.1038/s41416-021-01521-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Tan A, Kuo YF, Goodwin JS. Potential overuse of screening mammography and its association with access to primary care. Med Care. 2014;52(6):490–495. 10.1097/MLR.0000000000000115. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.