Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2018 Oct 1.
Published in final edited form as: Ann Epidemiol. 2017 Sep 20;27(10):654–658. doi: 10.1016/j.annepidem.2017.09.002

Absence of an anticipated racial disparity in interval breast cancer within a large healthcare organization

Garth H Rauscher a, Firas Dabbous b, Therese A Dolecek a, Sarah M Friedewald c, Katherine Tossas-Milligan a, Teresita Macarol b, W Thomas Summerfelt b
PMCID: PMC5682216  NIHMSID: NIHMS907338  PMID: 28964641

Abstract

PURPOSE

Compared to non-Latina (nL) white women, nL Black women are diagnosed with more aggressive breast cancers, which in turn should be more likely to go undetected on screening mammography and subsequently arise as interval breast cancer (IBC). We sought to estimate the extent of an anticipated racial disparity in IBC within a single, large health care organization.

METHODS

The present analyses focuses on 4,357 breast cancers diagnosed between 2001 and 2012 and within 18 months of a screening mammogram (N=714,218). We used logistic regression with model-based standardization (predictive margins) to estimate adjusted prevalence differences (PD) corresponding to a racial disparity in IBC.

RESULTS

Overall, prevalence of IBC within 18 months was 20.7 percent. Contrary to expectation, in patient-adjusted models, there was no IBC racial disparity (percentage point disparity = −2.1, 95% CI: −4.7, 2.6). However, when controlling for facility characteristics, including proportion nL Black patients, the model coefficient for the IBC disparity reversed sign and changed substantially (p<0.0001) and a racial disparity emerged (percentage point disparity = +5.1, 95% CI: −0.3, 9.9).

CONCLUSIONS

The sorting of patients by race across facilities appears to have mitigated an otherwise anticipated disparity in interval breast cancer. Possible explanations are discussed.

Keywords: Breast Neoplasms, African Americans, Caucasian Race, Mammography

INTRODUCTION

Mammography screening increases the chances of early detection for breast cancer and the balance of evidence suggests that it also reduces mortality from breast cancer (1–4). Although mammography screening has made a substantial contribution to reducing breast cancer morbidity and mortality, the sensitivity of mammography is inherently limited (varying from 68–93%), especially in young women and women with dense breast tissue (5). Therefore, some breast cancers among screened women are diagnosed as so-called interval breast cancer (IBC); that is, they arise between screens, as the result of a symptom or an indication on a breast physical examination.

Non-Hispanic (nL) black women are more likely than nL whites to be diagnosed with certain, more aggressive forms of breast cancer, including those that are high grade or negative for estrogen and progesterone receptors, and these tumors are more likely to evade detection with mammography (6–8). In addition to having more aggressive tumors, nL Black women also appear to be less likely to have access to the highest quality breast cancer screenings, according to studies of mammography access and quality conducted in Chicago (9–15). A recent analysis from the population-based Breast Cancer Care in Chicago study revealed that nL Black breast cancer patients were more likely than their nL White counterparts to report symptomatic awareness of their breast cancer despite a recent screening mammogram (10). For these reasons, it would be reasonable to expect an excess diagnosis of IBC among nL black women. In the present analysis, we sought to estimate the extent of an anticipated racial difference in IBC within a single large health care organization.

MATERIALS AND METHODS

The study received approval from the Institutional Review Boards at Advocate Health Care, University of Illinois at Chicago and the Illinois Department of Public Health. Data on mammography examinations came from a single large health care delivery organization with healthcare facilities throughout metropolitan Chicago. All sites are connected via a radiology information system developed by PenRad Technologies of Plymouth, MN. The PenRad database is used to enter all breast imaging results, biopsies and pathology results, and related data needed for Mammography Quality Standards Act (MQSA) compliance. For these analyses, a screening mammogram was defined as a bilateral mammogram with a description of screening in the radiology database, in women without a prior history of breast cancer, mastectomy, or breast implants, and without any imaging in the 9 months prior to the screen. We linked 714,218 screening examinations conducted between 2001–2010 to the population-based Illinois State Cancer Registry (ISCR) and identified 4357 breast cancers diagnosed within 18 months of a screen, between 2001 and 2012. The linkage was performed using probabilistic methods and the software Automatch, version 4 (1996) (16). Automatch is a highly flexible linkage software developed by Matchware Technologies, Inc. Through comparison with known breast cancers diagnosed from hospital tumor registries, approximately 99% of the breast cancers had a corresponding breast cancer record in ISCR.

Definition of interval breast cancer (IBC)

IBC was defined as an in-situ or invasive breast cancer diagnosed within 18 months following a screening mammogram (the “index” mammogram) with a negative finding (BIRADS 1, 2), but prior to a next screen. A breast cancer was defined as a screen-detected breast cancer if it was diagnosed within 18 months of a screening mammogram with a negative finding but ultimately diagnosed after a second screen with an abnormal finding.

Patient and clinical factors

Race/ethnicity was self-reported as nL White, nL Black, Latina, other, and unknown. Age at exam was defined in years and categorized as <40, 40–49, 50–59, 60–69, 70–79, and 80+. Body mass index (BMI) was categorized based on self-reported height and weight at screening exam as underweight (BMI<20) normal weight (20–24.99) over weight (25–29.99) obese (30–34.99) and morbidly obese (BMI>35). Breast density was recorded in the radiology database using the Breast Imaging Reporting and Data Systems (BIRADS) 4th edition categories of fatty, scattered fibroglandular, heterogeneously dense, and extremely dense. Family history of breast cancer was defined as none (no first or second degree relatives affected), weak (only second degree relatives affected), moderate (one first degree relatives over age 50 affected), and strong (multiple first degree relatives affected or at least one under age 50). Time between the two most recent screens prior to diagnosis was calculated as the difference in the date of the index screening mammogram and most recent prior screening mammogram and categorized as 9–18, 19–24, 25–36, 37–48, >48 months or first screen in the database. Presence or absence of a comparison film during interpretation of the index screen was available in the database. Two measures of socioeconomic status that were based on each woman’s census tract of residence (concentrated disadvantage and concentrated affluence) were also calculated (11).

Facility characteristics

We defined a variable to represent each screening facility’s proportion of patients that was nL black. Each mammography screening facility was as also categorized as either a stand-alone site, situated within a hospital that was not a comprehensive breast center, or that was situated within a comprehensive breast center. In addition, each screening exam was defined as either screen film (analog) or full field digital mammography (FFDM).

Tumor characteristics

Estrogen receptor and progesterone receptor status were each defined as negative or positive based on information from tumor registries. Likewise, histologic grade was defined as low, intermediate and high, and stage at diagnosis was categorized into American Joint Committee on Cancer (AJCC) categories of 0, I, II, III, and IV.

Statistical analyses

Ordinal and continuous variables were categorized for descriptive analyses of IBC predictors. We tabulated the distribution of patient clinical characteristics as well as facility and tumor characteristics separately for nL black and nL white patients. We also tabulated the percentage of breast cancers following screening mammography that were IBC by these same characteristics. In both instances, we obtained a p-value for differences in IBC (by race) from a Chi-Squared test (for nominal covariates) or from a test for trend for ordered covariates.

Next, a series of multivariable logistic regression models were conducted with IBC as the dependent variable. For these we restricted our sample to screening mammograms conducted on nL Black and nL White patients, of which there were 3810 breast cancers diagnosed within 18 months of a screening mammogram. Likelihood ratio tests were conducted to compare fit across nested models. Manual backwards selection (type 3 analysis using an alpha=0.10) was used to identify potentially important patient and clinical characteristics beyond age, to control for in all subsequent models (results not shown). Based on these results, breast density, BMI, time since last screen, availability of a comparison film and type of mammogram (analog or digital) were identified as covariates in addition to age and family history (included a-priori) for the mediation analyses described below.

Mediation analyses

We used two approaches in order to examine the potential mediation of a racial disparity in IBC. First, we conducted logistic regression with model-based standardization (predictive margins) in order to estimate the prevalence difference and 95% confidence interval (CI) in IBC prevalence by race (17, 18). Separate models controlled for (1) age, (2) age and patient characteristics, and (3) age, patient, and facility characteristics (facility type and proportion nL black). We used the predictive margins from these models to estimate a series of average controlled direct associations in the form of prevalence differences to estimate what the racial disparity in IBC prevalence might be if the distributions of patient and/or facility characteristics were equalized within race (11). We expressed associations for the black-white disparity as prevalence (percentage-point) differences that were on a scale that is more relevant for public health interpretation than risk ratios or odds ratios are (19, 20). We estimated bias-corrected, bootstrapped 95% confidence intervals for these measures.

Second, we compared rescaled coefficients from probit regression models representing the racial difference in IBC prevalence before and after adjusting for facility characteristics (21). The method equalizes the variances and allows for a fair comparison of the magnitude of coefficients between a “reduced” model (without a set of hypothesized mediators) and a “full” model (with the mediators of interest). The p-value from the difference in the race coefficient from reduced and full models was used to indicate statistical evidence for mediation by facility characteristics on the racial disparity in IBC (18). All analyses were conducted using Stata statistical software, version 12 (Stata Corp, College Station, TX). All p-values are two-sided.

RESULTS

Racial differences in patient, facility and tumor characteristics

The distribution of age was similar for nL Black and nL White patients. NL Black patients were considerably more likely to reside in more disadvantaged and less affluent census tracts. Compared with nL blacks, nL white patients were slightly more likely have dense breasts, more likely to have obtain a digital as opposed to an analog mammogram and to have their screen interpreted in the context of a comparison film, and considerably more likely to be screened at hospital sites (Table 1). NL Black patients were more likely than their nL White counterparts to be diagnosed with more aggressive tumors that were ER negative, PR negative and higher grade. There was no apparent disparity in stage at diagnosis among this sample of screened women (Table 1).

Table 1.

Distribution of demographic/clinical, facility, and tumor characteristics by patient race for 3810 nL white and nL Black breast cancer patients screened from 2001–2010 and diagnosed between 2001–2012.

Characteristics nL White
nL Black
No. % No. % P- Value
Age group 0.005
 40 – 49 584 22 204 17
 50 – 59 743 28 333 28
 60–69 724 28 372 32
 70 – 79 581 22 269 23
BMI <0.0001
 Under weight 66 4 11 1
 Normal weight 426 28 92 12
 Over weight 494 33 214 29
 Obese 328 22 207 28
 Morbidly Obese 205 13 220 30
Family history
 None 1469 56 713 61
 Weak 510 19 176 15
 Moderate 431 16 202 17
 Strong 222 8 87 7
Breast Density 0.003
 Fatty 244 9 83 7
 Scattered 926 35 511 43
 Heterogeneous 1235 47 529 45
 Dense 225 9 53 4
Time since last screen
 9 – 14.9 months 945 36 400 34
 15 – 20.9 months 288 11 182 15
 21 – 29.9 months 198 8 122 10
 30+ months 1201 46 474 40
Comparison Film <0.0001
 None 405 15 305 26
 Yes 2227 85 873 74
Imaging Modality <0.0001
 Analog 1348 51 818 69
 Digital 1284 49 360 31
Screened at hospital site <0.0001
 No 932 35 720 61
 Yes 1700 65 458 39
Comprehensive breast center <0.0001
 No 1787 68 1064 90
 Yes 845 32 114 10
ER Status <0.0001
 Negative 316 12 216 18
 Positive 1886 72 770 65
 Missing 430 16 192 16
Tumor Grade <0.0001
 Low 622 24 200 17
 Moderate 1092 41 490 42
 High 752 29 401 34
 Missing 166 6 87 7
Stage at Diagnosis
 0 678 26 326 28
 I 1225 47 483 41
 II 446 17 234 20
 III 233 9 107 9
 IV 35 1 13 1
 Missing 15 1 15 1
Open in a new tab

P-values >0.01 are suppressed

IBC associations with patient, facility, and tumor characteristics

Overall, prevalence of IBC within 18 months was 20.7 percent. Table 2 presents associations with respect to prevalence of IBC. Contrary to expectation, prevalence of IBC was slightly lower for nL Black vs. nL white patients. As expected, younger age and denser breasts were each associated with increased prevalence of IBC. Prevalence of IBC was greater for patients with a shorter time between screens and for patients whose index screen was interpreted with the aid of a comparison film (Table 2). More aggressive tumors (ER/PR negative and higher grade) were more likely to be diagnosed as IBC. In addition, tumors with a lobular histology were more likely to be diagnosed as IBC. IBC was strongly associated with later stage at diagnosis (Table 2). Contrary to expectation, prevalence of IBC was greater at comprehensive breast centers compared to non-academic hospitals and standalone sites. IBC prevalence was greatest for patients screened at facilities with the lowest proportion of nL Black patients (23%) and IBC prevalence was lowest for patients screened at facilities with the highest proportion of nL Black patients (17%) (p<0.0001) (Table 2).

Table 2.

Patient/clinical, facility, and tumor characteristics and associations with interval breast cancer (IBC) among 3810 nL white and nL Black breast cancer patients and 547 other patients screened from 2001–2010 and diagnosed between 2001 – 2012.

Characteristics IBC within 18 mos. of screen
No. % P-Value
Race/ethnicity 0.09
 nL White 2632 21
 nL Black 1178 19
 Other/missing 547 24
Age group 0.0003
 40–49 935 25
 50–59 1244 22
 60–69 1245 18
 70–79 933 20
BMIa 0.001
 Under weight 87 26
 Normal weight 596 23
 Over weight 799 19
 Obese 585 19
 Morbidly Obese 448 16
Family history
 None 2570 21
 Weak 764 22
 Moderate 681 20
 Strong 342 20
Breast Density <0.0001
 Fatty 388 10
 Scattered 1607 17
 Heterogeneous 2019 24
 Dense 339 34
Time since last screen <0.0001
 9–14.9 months 1504 28
 15–20.9 months 520 24
 21–29.9 months 367 19
 30+ months 1966 15
Comparison Film <0.0001
 None 836 12
 Yes 3521 23
Imaging Modality <0.0001
 Analog 2477 18
 Digital 1880 26
Screening site <0.0001
 Stand alone 1952 21
 Non-academic hospital 1303 18
 Comprehensive Breast Centers 1102 26
Proportion nL Black <0.0001
 Low 3071 23
 Moderate 294 20
 High 992 17
ER Status <0.0001
 Negative 604 30
 Positive 3016 21
Tumor Grade 0.0005
 Low 938 20
 Moderate 1797 19
 High 1325 25
Open in a new tab

P-values >0.10 are suppressed

a

BMI is missing for roughly half the sample.

Mediation of the black-white disparity in IBC by facility characteristics

In a baseline model adjusting for age (and before adjusting for patient or facility characteristics), nL black patients were roughly two percentage points less likely than nL White patients to experience an interval breast cancer; after adjusting for facility characteristics nL black patients appeared to be roughly five percentage points more likely than nL White patients to experience an interval breast cancer (Table 3). Before adjusting for facility characteristics there was no disparity in the prevalence of interval breast cancer, but after adjusting for facility characteristics a disparity emerged. When we repeated these analyses for patients with data on BMI (which reduced our sample from N=3778 to N=2263 patients) results were qualitatively similar, except that the disparity began to emerge with adjustment for patient factors including BMI, and again the disparity was magnified with additional adjustment for facility level factors.

Table 3.

The role of measured facility characteristics versus individual facilities in mediating the racial disparity (nL Black vs. nL White) in the prevalence of interval breast cancer diagnosed within 18 months of a screening mammogram.

N PD (95% CI)1 (P-value)3
All available data
 Baseline adjust for age 3788 – 2.1 (−4.7, 2.6)
 Add patient characteristics except BMI4 3788 0.0 (−3.0, 3.4) <0.0001
 Add facility-level factors5 3788 4.5 (−1.0, 9.7) <0.0001
 Add both patient and facility-level factors 3788 5.1 (0.3, 9.9) <0.0001
For patients with data on BMI
 Baseline adjust for age 2238 – 0.1 (−3.3, 2.3)
 Add patient characteristics except BMI4 2238 1.6 (−1.8, 5.0) <0.0001
 Add patient characteristics including BMI4 2238 2.8 (−0.11, 6.8) <0.0001
 Add facility-level factors5 2238 8.7 (2.9, 14.9) <0.0001
 Add both patient and facility-level factors 2238 9.5 (4.4, 17.7) <0.0001
Open in a new tab
1

PD Prevalence difference for the Black White disparity in interval breast cancer within 18 months of a screening mammogram, estimated via logistic regression with marginal standardization, 95% CI via bias-corrected bootstrap methods.

3

P-value from a test of the difference in rescaled coefficients compared to the baseline model.

4

BMI, analog vs. digital mammogram, availability of a comparison film, breast density and time since last screen.

5

Facility type (stand-alone, hospital, non-comprehensive, or comprehensive breast center) and proportion of patients that were nL Black.

We conducted exploratory analyses to examine the hypothesis that nL black patients were being screened disproportionately at facilities with higher false positive rates which in turn might be associated with lower interval cancer prevalence. We aggregated records in the patient-level dataset by facility and calendar year and calculated the mean interval cancer prevalence, mean false positive rate, and mean cancer detection rate for each combination of facility and calendar year. We compared false positive rates across facilities and calendar years by patient race, and created scatter plots to compare cancer detection rates, false positive rates, and interval breast cancer prevalence across facilities and calendar years (Figure 1). NL Black patients were more likely than nL Whites to be screened at facilities and calendar years with higher false positive rates (Panel A). Higher false positive rates, in turn, were strongly associated with higher cancer detection rates (Panel B) which in turn were associated with lower interval cancer prevalence (Panel B).

FIGURE 1.

FIGURE 1

Open in a new tab

All data are aggregated at the facility and calendar year level (one observation for wach combination of facility and calendar year). Panel A: Boxplot comparing false positive rates for nL WhIte and nL Black patients. Panel B: Scatter plot where each dot represents a aggregate data for cancer detection rate and false positive rate for a specific facility and calendar year, along with a best fitting linear projection from a regression of cancer detection rate regressed on false positive rate. Panel C: Scatter plot where each dot represents a aggregate data for interval breast cancer prevalence and cancer detection rate for a specific facility and calendar year, along with a best fitting linear projection from a regression of interval breast cancer prevalence regressed on cancer detection rate.

DISCUSSION

We anticipated that, since nL Black patients tend to be diagnosed with more aggressive forms of breast cancer, including those that are high grade or negative for estrogen and progesterone receptors, they would likewise be more likely than nL whites to be diagnosed with interval breast cancer. In addition, the expectation of a racial disparity was based on prior research conducted in Chicago showing that, compared to their nL white counterparts, nL Black women may not have the same level of access to high quality breast cancer screening and diagnostic services, which might lead to an exacerbation of expected disparities in the prevalence of interval breast cancer (9–15). Within this single, large healthcare organization a racial disparity in interval breast cancer was not readily apparent. However, an underlying racial disparity might have been masked by the way in which patients of different racial/ethnic backgrounds were distributed across facilities, perhaps due to geographic proximity or referral. A variable representing the proportion of nL Black patients screened at each facility proved to be strongly associated with prevalence of IBC and appeared to account for the lack of an observed racial disparity in IBC; that, is, when we included this variable in our models of IBC an otherwise absent disparity emerged. Patients screened at facilities with higher proportions of nL Black patients were more likely to avoid an interval cancer, but it is not clear what specific (unmeasured) aspects of either patients or facilities might account for this. NL White patients were more likely to be screened at a comprehensive breast center which in turn (contrary to expectation) was associated with increased prevalence of IBC, despite the higher quality screening imaging and interpretation that we would anticipate at comprehensive breast centers (22, 23). Perhaps there is an increased focus on reducing false positives at comprehensive sites, which might come with the unanticipated consequence of somewhat higher interval cancer prevalence. In additional exploratory analyses using measures aggregated at the facility and calendar year, we found that NL Black patients were more likely than nL Whites to be screened at facilities and calendar years with higher false positive rates which in turn was associated with higher cancer detection rates and lower interval cancer prevalence.

NL Black women were less likely to have a comparison film available at screening interpretation which in turn was associated with lower prevalence of interval breast cancer. Perhaps the benefit of having a comparison film with respect to reducing callback rates and false positive screens was somewhat offset by an increased tendency to overlook lesions that might later arise as breast cancer, and therefore nL Black women may have experienced a lower likelihood of interval breast cancer at the expense of a higher likelihood of a false positive screen. During this time period in this organization nL Black women were also more likely to have been screened with an analog as opposed to a digital machine and being screened with an analog machine was associated with lower interval breast cancer prevalence.

Our study has certain limitations. While the sample of women included in this study did encompass the entire mammography screening population within a single, large healthcare organization, including multiple urban and suburban, hospital and non-hospital based locations across metropolitan Chicago, it was not a population-based sample. We lacked detailed measures of individual facilities that could potentially be markers of the quality of screening that might impact IBC prevalence, and had to rely on broad institution type and patient mix with respect to race.

In conclusion, the sorting of patients by race across facilities appears to have mitigated an otherwise anticipated disparity in interval breast cancer. NL Black patients were more likely than nL Whites to be screened at facilities with higher false positive rates which in turn were associated with higher cancer detection rates and less interval breast cancer. This appears to partially explain the lack of an anticipated disparity in IBC.

Acknowledgments

FUNDING

This work was supported by a grant from the Agency for Health Research and Quality to the University of Illinois at Chicago (R01 HS018366-01A1) and a grant from the National Cancer Institute to Group Health Cooperative (1P01CA154292-01A1)

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Conflict of Interest and Authorship Conformation Form

Please check the following as appropriate:
  • ✓
    All authors have participated in (a) conception and design, or analysis and interpretation of the data; (b) drafting the article or revising it critically for important intellectual content; and (c) approval of the final version.
  • ✓
    This manuscript has not been submitted to, nor is under review at, another journal or other publishing venue.
  • ✓
    The authors have no affiliation with any organization with a direct or indirect financial interest in the subject matter discussed in the manuscript

References

  • 1.Nelson HD, Tyne K, Naik A, Bougatsos C, Chan BK, Humphrey L, et al. Screening for breast cancer: an update for the U.S. Preventive Services Task Force. Annals of Internal Medicine. 2009;151(10):727–37. W237–42. doi: 10.1059/0003-4819-151-10-200911170-00009. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Berry DA, Cronin KA, Plevritis SK, Fryback DG, Clarke L, Zelen M, et al. Effect of screening and adjuvant therapy on mortality from breast cancer. New England Journal of Medicine. 2005;353(17):1784–92. doi: 10.1056/NEJMoa050518. [DOI] [PubMed] [Google Scholar]
  • 3.Shen Y, Yang Y, Inoue LY, Munsell MF, Miller AB, Berry DA. Role of detection method in predicting breast cancer survival: analysis of randomized screening trials. Journal of the National Cancer Institute. 2005;97(16):1195–203. doi: 10.1093/jnci/dji239. [DOI] [PubMed] [Google Scholar]
  • 4.Smith RA, Duffy SW, Gabe R, Tabar L, Yen AM, Chen TH. The randomized trials of breast cancer screening: what have we learned? Radiologic Clinics of North America. 2004;42(5):793–806. doi: 10.1016/j.rcl.2004.06.014. [DOI] [PubMed] [Google Scholar]
  • 5.Mandelson MT, Oestreicher N, Porter PL, White D, Finder CA, Taplin SH, et al. Breast density as a predictor of mammographic detection: comparison of interval- and screen-detected cancers. Journal of the National Cancer Institute. 2000;92(13):1081–7. doi: 10.1093/jnci/92.13.1081. [DOI] [PubMed] [Google Scholar]
  • 6.Miller BA, Hankey BF, Thomas TL. Impact of sociodemographic factors, hormone receptor status, and tumor grade on ethnic differences in tumor stage and size for breast cancer in US women. American Journal of Epidemiology. 2002;155(6):534–45. doi: 10.1093/aje/155.6.534. [DOI] [PubMed] [Google Scholar]
  • 7.Chlebowski RT, Chen Z, Anderson GL, Rohan T, Aragaki A, Lane D, et al. Ethnicity and breast cancer: factors influencing differences in incidence and outcome. Journal of the National Cancer Institute. 2005;97(6):439–48. doi: 10.1093/jnci/dji064. [DOI] [PubMed] [Google Scholar]
  • 8.Joslyn SA. Hormone receptors in breast cancer: racial differences in distribution and survival. Breast Cancer Research & Treatment. 2002;73(1):45–59. doi: 10.1023/a:1015220420400. [DOI] [PubMed] [Google Scholar]
  • 9.Molina Y, Silva A, Rauscher GH. Racial/ethnic disparities in time to a breast cancer diagnosis: the mediating effects of healthcare facility factors. Medical Care. 2015;53:872–878. doi: 10.1097/MLR.0000000000000417. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Mortel M, Rauscher GH, Murphy AM, Hoskins K, Warnecke RB. Racial and Ethnic disparity in symptomatic breast cancer awareness despite a recent screen: the role of tumor biology and mammography facility characteristics. Cancer Epidemiol Biomarker Prev. 2015;24(10):1599–606. doi: 10.1158/1055-9965.EPI-15-0305. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Rauscher GH, Campbell RT, Wiley EL, Hoskins K, Stolley M, Warnecke RB. Socioeconomic position and reproductive factors mediate racial and ethnic disparities in estrogen/progesterone receptor negative breast cancer. American Journal of Epidemiology. 2015 doi: 10.1093/aje/kwv226. In Press. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Rauscher GH, Allgood KL, Whitman S, Conant E. Disparities in screening mammography services by race/ethnicity and health insurance. Journal of Women’s Health. 2012;21(2):154–60. doi: 10.1089/jwh.2010.2415. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Rauscher GH, Conant EF, Khan JA, Berbaum ML. Mammogram image quality as a potential contributor to disparities in breast cancer stage at diagnosis: an observational study. BMC Cancer. 2013;13:208. doi: 10.1186/1471-2407-13-208. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Rauscher GH, Khan JA, Berbaum ML, Conant EF. Potentially missed detection with screening mammography: does the quality of radiologist’s interpretation vary by patient socioeconomic advantage/disadvantage? Annals of Epidemiology. 2013;23(4):210–4. doi: 10.1016/j.annepidem.2013.01.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Rauscher GH, Murphy AM, Orsi JM, Dupuy DM, Grabler PM, Weldon CB. Beyond the mammography quality standards act: measuring the quality of breast cancer screening programs. AJR American Journal of Roentgenology. 2014;202(1):145–51. doi: 10.2214/AJR.13.10806. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Leonard DR, Rariden RL, Beccue B, Shen T. Attempts to Minimize Manual Review During Registry Deduplication. Journal of Registry Management. 2006;33(1):17–23. [Google Scholar]
  • 17.Greenland S. Model-based estimation of relative risks and other epidemiologic measures in studies of common outcomes and in case-control studies. American Journal of Epidemiology. 2004;160(4):301–5. doi: 10.1093/aje/kwh221. [DOI] [PubMed] [Google Scholar]
  • 18.Ahern J, Hubbard A, Galea S. Estimating the effects of potential public health interventions on population disease burden: a step-by-step illustration of causal inference methods. American Journal of Epidemiology. 2009;169(9):1140–7. doi: 10.1093/aje/kwp015. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Poole C. A history of the population attributable fraction and related measures. Annals of Epidemiology. 2015;25(3):147–54. doi: 10.1016/j.annepidem.2014.11.015. [DOI] [PubMed] [Google Scholar]
  • 20.Poole C. On the origin of risk relativism. Epidemiology. 2010;21(1):3–9. doi: 10.1097/EDE.0b013e3181c30eba. [DOI] [PubMed] [Google Scholar]
  • 21.Karlson KB. Decomposing primary and secondary effects: A new decomposition method. Research in Social Stratification and Mobility. 2011;29(2):221–37. [Google Scholar]
  • 22.Beam CA, Conant EF, Sickles EA. Association of volume and volume-independent factors with accuracy in screening mammogram interpretation. Journal of the National Cancer Institute. 2003;95(4):282–90. doi: 10.1093/jnci/95.4.282. [DOI] [PubMed] [Google Scholar]
  • 23.Skinner KA, Helsper JT, Deapen D, Ye W, Sposto R. Breast cancer: do specialists make a difference? Annals of Surgical Oncology. 2003;10(6):606–15. doi: 10.1245/aso.2003.06.017. [DOI] [PubMed] [Google Scholar]

RESOURCES