Determinants of Breast Cancer Treatment Delay Differ for African American and White Women

African American women have a higher breast cancer mortality compared with White women, even after accounting for clinical and prognostic factors (1–4) and socioeconomic characteristics (5). Timeliness of treatment has been used as an indicator of quality of care (6, 7) and may contribute in part to these persistent disparities. African American women experience greater delays in care than White women at multiple points along the treatment pathway from detection to medical consultation/diagnosis (“diagnostic delay”; refs. 7–12) and from diagnosis to the initiation of treatment (“treatment delay”; refs. 7–9, 13–17). Although the majority of studies show that African American women are more likely than White women to experience treatment delay (8, 14, 15, 17–20), not all studies find differences between these groups (21–23).

The impact of socioeconomic characteristics has been heavily investigated, but does not fully explain racial disparities in timeliness of care. A study on Medicare beneficiaries found that African American women were more likely than White women to experience delays between initial consultation (a diagnostic imaging procedure or consultation for symptoms) and diagnosis, as well as between diagnosis and treatment (9). In a Washington, DC cohort, African American women were more likely than White women to experience delay between the identification of a suspicious finding and diagnostic resolution even among women with the same type of insurance coverage (private or government; ref. 10). Among low-income, uninsured women enrolled in the National Breast and Cervical Cancer Early Detection Program (NBCCEDP), African American women are more likely to experience diagnostic delay and treatment delay (7).

Limitations in the assessment of socioeconomic characteristics may affect the validity and interpretation of the results for many studies conducted to date. First, detailed socioeconomic data are often unavailable in investigations with large study populations (24). Area-level data (e.g., census tracts, zip codes) are used as a proxy for individual-level income and education (8, 9, 15, 25–29). These measures may be unreliable when there is marked heterogeneity within the area being analyzed (30). In addition, they may not adequately control for confounding as studies suggest that area-level and individual-level characteristics independently impact breast cancer outcomes (31, 32). Second, studies often combine persons with Medicare and Medicaid coverage into a single category (10, 13, 14). Results for this heterogeneous category are difficult to interpret, particularly given the marked distinctions in breast cancer outcomes for these groups (15, 16, 33).

Identifying additional determinants of treatment delay will improve our understanding of racial disparities and is critical for developing interventions and policy to ensure timely care. Investigations based solely on administrative, cancer registry, or medical record data are fairly limited in the types of information they collect, and thus are less amenable to discovering novel determinants of delay (7, 28). The goals of this study were to identify determinants of breast cancer treatment delay and to determine whether disparities in treatment delay exist between White and African American women. We use data from a population-based study to assess the association of demographic and socioeconomic characteristics, indicators of healthcare access, clinical factors, and measures of emotional and functional well-being with treatment delay. By combining medical record and patient-reported data, we were able to assess several factors that are not typically evaluated in other investigations and to obtain individual-level socioeconomic data.

The Carolina Breast Cancer Study (CBCS) phase III is an ongoing population-based study of breast cancer in North Carolina. The design is similar to earlier phases (34, 35), except that it is a case-only study and has a larger recruitment area (44 counties). Eligible participants are (i) 20 to 74 years old, (ii) North Carolina residents at the time of diagnosis, and (iii) have incident, pathologically confirmed invasive primary breast cancer. Women with a previous diagnosis of invasive breast cancer are excluded. A random sample of eligible women is selected from the following strata: (i) African Americans <50 years old, (ii) African Americans ≥50 years old, (iii) non-African Americans <50 years old, and (iv) non-African Americans ≥50 years old. The sampling fractions are 100%, 60%, 40%, and 15%, respectively. At baseline, participants are interviewed by a nurse, complete a quality-of-life questionnaire, and provide written consent for medical record requests.

The present study uses abstracted medical record data and baseline nurse-administered interview and questionnaire data for women diagnosed between May 1, 2008 and January 29, 2010 (N = 771). Participants were non-Hispanic White (“White”) or non-Hispanic African American (“African American”), received either surgery or neoadjuvant chemotherapy and/or hormone therapy (“neoadjuvant therapy”) as their first course of treatment, had a known treatment date, and were diagnosed on the basis of a core needle biopsy (CNB) before treatment. Participants who received preoperative chemotherapy or hormonal therapy, or pre- and postoperative chemotherapy or hormone therapy were classified as having received neoadjuvant therapy as their first course of treatment. Hispanics and other racial groups were not analyzed because of small numbers (<3% for both). The final study population consisted of 601 women (Fig. 1). This study was approved by the Office of Human Research Ethics at the University of North Carolina at Chapel Hill (Chapel Hill, NC).

The main exposure, race, was obtained from self-report. We calculated treatment delay (main outcome) as the time in days between the date of the CNB used to diagnose invasive disease and the initiation of the first course of treatment. Treatment delay was dichotomized as more than 30 days (“delay”) or 30 days or less. Although this is a commonly used threshold (7, 9, 16, 17), the clinically relevant delay for first course of treatment is unknown. Additional details on the study variables are provided in the Supplementary Data.

The percentages shown in Tables 1 and 2 were weighted to obtain North Carolina population estimates (hereafter referred to as the “overall population”). The sample sizes shown in the tables are unweighted. The study design was accounted for in tests of associations between categorical variables. Logistic regression was used to estimate the odds ratios (ORs) and 95% confidence interval (95% CI) for the association between each exposure and delay. Separate models were developed for each exposure (race and characteristics of delay identified in bivariate analyses) to account for the different sets of explanatory variables necessary to control for confounding (36, 37). On the basis of the literature, the following explanatory variables were included in the adjusted models to estimate the direct effect of race (White or African American) on delay: age (20–39, 40–49, 50–64, or 65–74); income (≤$20,000, $20,000–$30,000, or >$30,000); insurance coverage (private, Medicare, Medicaid, or none); education (0–12 years, but no high school degree; high school graduate; some college; technical or business school; or college degree or higher); lost a job due to one's diagnosis (yes or no); American Joint Committee on Cancer (AJCC) disease stage (I, IIA, IIB, or III/IV); symptoms (yes or no); first treatment/reconstruction [breast conserving surgery (BCS), mastectomy without reconstruction, mastectomy with reconstruction, or neoadjuvant therapy]; and marital status (married or unmarried). We estimated the total effect of the following characteristics on delay: the number of people supported by the household income (“household size”; explanatory variables: age, race, and marital status), losing a job due to one's diagnosis (explanatory variables: age, race, and education), and first treatment/reconstruction (explanatory variables: age, race, education, income, insurance, and disease stage). All variables were treated as categorical. An interaction term between age and race was included in all models to account for the study design. The study population rather than the overall population was used in the models to ensure a sufficient number of women of both races for each age group. All P values were two-sided and P < 0.05 was considered statistically significant. Statistical analyses were conducted using SAS 9.2 (SAS Institute Inc.).

The contact rate for the study (number of women selected for the study − number of women who could not be located or did not respond) was 95.4%. The cooperation rate (number of women who completed interviews/number of women contacted and eligible) was 80.2%. The overall response rate [number of completed interviews/(number of women selected − number ineligible or deceased); ref. 38] was 75.9%. The median time elapsed between diagnosis and the baseline interview was 5.1 months [interquartile range (IQR), 4.0–6.2 months]. The median time from diagnosis until treatment was 27.0 days (IQR, 18.0–37.0 days) for the study population and 26.2 days (IQR, 16.3–36.1 days) for the overall population. The majority (84.2%) of women in the overall population had a CNB before treatment. There was no association between receipt of a CNB and race (P = 0.39) in the overall population (or study population). There was a significant association between receipt of a CNB and income (P = 0.047) in the overall population: 93.3% of women with an income of more than $100,000 had a CNB versus 79.2% to 81.6% of women with lower incomes.

As shown in Table 1, delay was significantly associated with household size (P = 0.011), losing a job due to one's diagnosis (P < 0.01), and immediate reconstruction after mastectomy (P < 0.01). Women with smaller households were more likely than women with larger households to experience delay: 35.2% to 47.3% for households of 3 or less people versus less than 30% for households of more than 3 people. Women who lost a job due to their diagnosis and who underwent immediate reconstruction were nearly twice as likely to experience delay compared with women who did not (73.6% vs. 37.8%, and 64.9% vs. 35.6%, respectively). The prevalence of women in the overall population with households of 3 or less people, who lost a job due to their diagnosis, and who underwent immediate reconstruction was 17.0%, 3.5%, and 10.6%, respectively. The latter value is the product of the immediate reconstruction and mastectomy rates. African American women were slightly more likely than White women to experience delay (43.4% vs. 38.4%; P = 0.24). None of the other characteristics were significantly associated with delay.

Few women experienced a delay of more than 60 days (2.6%; data not shown). Race was strongly associated with delays of this length, with African American women being more than 3 times as likely as White women to experience delay (6.0% vs. 1.7%; P = 0.013). For most categories, less than 5% of women experienced a delay of more than 60 days. Women in the following categories were exceptions: underwent immediate reconstruction (12.9%), were unable to see a doctor in the past 10 years due to a lack of transportation (9.3%), completed technical or business school (6.3%), underwent mastectomy (6.1%), had not accepted their illness at all (6.1%), had Medicaid coverage (5.5%), and had a household of 4 people (5.3%).

African American women differed markedly from White women for nearly every demographic and socioeconomic characteristic, measure of healthcare access, and clinical factor (Table 2). The following characteristics were exceptions: working since diagnosis (P = 0.69), family history (P = 0.35), first course of treatment (P = 0.46), and immediate reconstruction (P = 0.079). Race was not associated with the measures of emotional and functional well-being: P = 0.67 and 0.26 for the degree of satisfaction with their coping and acceptance of their illness, respectively. Among the characteristics associated with delay, African American women were more than twice as likely as White women to lose a job due to their diagnosis (6.6% vs. 2.7%), slightly more likely to have a household of 3 or less people (84.6% vs. 82.5%), and less likely to undergo immediate reconstruction (25.2% vs. 37.9%).

The stratified data (Table 2) revealed that the determinants of delay are not equivalent for White and African American women. The only determinant common to both groups was immediate reconstruction (P < 0.01 for each group). Household size and losing a job due to one's diagnosis were significantly associated with delay among White women (P = 0.027 and P < 0.01, respectively), whereas the first course of treatment was significantly associated with delay among African American women (P < 0.01).

Although race was not associated with delay in the aggregated data, the stratified data revealed racial disparities for women with similar characteristics (e.g., same educational level). African American women were more likely than White women to experience delay for most characteristics. For instance, 55.6% of African American women with stage IIB disease experienced delay versus 36.5% of White women with stage IIB disease. The frequency of delay for African American women exceeded the frequency for White women by more than 30% for the following categories: detection by a method other than a routine mammogram, clinical breast examination, or self- or spouse-detection (70.0% for African American women vs. 33.3% for White women), no insurance coverage (48.9% for African American women vs. 15.3% for White women), households of 4 people (53.5% for African American women vs. 23.2% for White women), and undergoing immediate reconstruction (92.5% for African American women vs. 60.6% for White women).

Losing a job due to one's diagnosis was associated with the highest probability of delay among White women. It was also the only characteristic in which the frequency of delay among White women was more than 30% higher than for African American women: 87.2% versus 52.6%. Having no insurance coverage was associated with the lowest probability of delay among White women (15.3%). Age (P = 0.067) and symptoms (P = 0.051) also showed evidence of an association with delay among White women. Women ages 50 to 64 years and without symptoms tended to be more likely to experience delays. Undergoing immediate reconstruction was associated with the highest probability of delay among African American women, whereas having less than a high school degree was associated with the lowest probability of delay (26.4%). Income (P = 0.090) showed evidence of an association with delay among African American women, but there was no clear trend among categories.

In the fully adjusted models, women with 2-person households were more likely than single-person households to experience delay (OR, 2.08; 95% CI, 1.19–3.63; Table 3). The probability of delay decreased with increasing household size for households of 2 or more people. Women who lost a job due to their diagnosis were more likely to experience delay compared with women who did not (OR, 2.19; 95% CI, 1.00–4.81), although the result was not significant (P = 0.050). Women who underwent mastectomy with immediate reconstruction (OR, 6.18; 95% CI, 3.27–11.68) and neoadjuvant therapy (OR, 2.06; 95% CI, 1.16–3.66) were more likely than women who received BCS to experience delay.

Racial disparities in delay were significant among women younger than 50 years (Table 4). The disparity was largely explained by the low likelihood of delay among younger White women. African American women were more than 3 times as likely as White women to experience delay among women 20 to 39 years old (OR, 3.34; 95% CI, 1.07–10.38) and 40 to 49 years old (OR, 3.40; 95% CI, 1.76–6.54). Among White women, delays were less likely for 20- to 39-year-old and 40- to 49-year-old women (OR, 0.32; 95% CI, 0.13–0.80 and OR, 0.38; 95% CI, 0.20–0.73, respectively) relative to women 50 to 64 years old. The likelihood of delay for African American women did not differ significantly based on age.

This study found that African American women were more likely than White women to experience delay among younger age groups (<50 years), but not among older age groups. This disparity was not evident in the overall population, as we found no association between race and delay in the aggregate data. Household size, losing a job due to one's diagnosis, and immediate reconstruction were associated with delay in the overall population and among White women, respectively. Among African American women, who were a minority in the overall population, immediate reconstruction and first course of treatment were associated with delay. The adjusted models showed that women with 2-person households experienced greater delay than women with other household sizes and women who had mastectomy with immediate reconstruction experienced greater delay than women who received other treatments.

It is unclear why having a 2-person household was associated with delay based on the available data. Further investigation is needed to understand this finding. Increased delay among women who lose a job due to their diagnosis may be related to a loss of employer-based insurance coverage, greater financial constraints, or an unsupportive work environment. We are unaware of any quantitative studies of treatment delay that evaluate employment changes. Losing a job may impact delay only among White women because they are more likely to have private coverage (1, 29, 39, 40). African American women were more likely than White women to lose a job even though the frequency of working since one's diagnosis was comparable for both racial groups. This result is in agreement with a study that found that African American women were more likely than White women, to stop working (OR, 3.0; 95% CI, 1.3–6.7) or miss work for more than 1 month (OR, 3.0; 95% CI, 1.2–7.4), respectively, compared with missing work for 1 month or less (41). African American women were also less likely than White women to be employed 18 months following diagnosis (OR, 0.35; 95% CI, 0.18–0.68; ref. 42).

The additional time necessary for consultation and coordination of the schedules of the plastic surgeon and primary surgeon may explain the increased delay associated with immediate reconstruction. This is not the first study to report increased delay associated with this procedure (19, 43), but most studies on treatment delay do not consider this factor. Immediate reconstruction was the only factor associated with delay for both racial groups. African American women were less likely than White women to undergo immediate reconstruction, but more likely to experience delays if they underwent this procedure. Although not significant, our finding that African American women are less likely to undergo immediate reconstruction has been observed in several studies (44–46). First course of treatment was associated with delay only among African American women and was not explained by differences in the types of treatment received. African American and White women may differ in this determinant because they receive care at different types of healthcare facilities as a result of insurance status and income (16, 47, 48), residential segregation (49, 50), and urban/rural residence (48). Healthcare facility characteristics are known to affect treatment delay (15, 17). Immediate reconstruction rates also vary based on the healthcare facility (51).

It is challenging to compare the frequency of treatment delay across studies due to variations in study design, recruitment criteria, and study population characteristics. The start point of the treatment delay period has been defined in various ways, including the date of first clinical confirmation (15, 23), a suspicious finding (15), and pathologic diagnosis (7). Eligibility restrictions based on disease stage (15, 21, 43) and type of treatment received [e.g., one type (16, 43, 52) vs. all types (14, 47)] also limit comparability. Recruitment of women from a specific healthcare facility or set of facilities (e.g., single vs. multiple, public/public safety net vs. private; refs. 16, 23, 53) and a specific program (22, 52) may lead to marked differences in population characteristics. The frequency of delay (>30 days) reported in the literature (7–9, 14–18, 52) ranges from 21.8% (7) to 68.9% (16). Our result (39.4%) is similar to the results from 2 different national studies (15, 17) that obtained values of 34.9% and 42.6%, respectively.

The frequency of delay reported in the literature for African American and White women (7–9, 14, 16–18) ranges from 18.7% (14) to 70.8% (16) and 4.7% (14) to 56.1% (16), respectively. Our results (43.3% and 38.4% for African American and White women, respectively) are fairly similar to those from a national study that reported values of 53.0% and 40.4%, respectively (17). Delays of more than 60 days were uncommon in this population, but are more frequent in studies focusing on women who are uninsured, have Medicaid coverage, or have low incomes (23, 54). Our finding that African American women are more likely to experience a delay of more than 60 days is consistent with other studies (17, 18). Because the determinants of delay identified in bivariate analysis are highly dependent on the predominant study population characteristics, this may explain the conflicting literature on racial disparities in treatment delay. Studies that do not report an association between race and delay tend to restrict the study population based on socioeconomic characteristics, disease stage, or the healthcare facility at which they receive care (21–23, 52). Differences in the determinants of treatment delay and racial disparities in delay are likely diminished for more homogeneous populations.

The findings of this study suggest the need to focus on well-defined populations when using treatment delay to make comparisons between groups, monitor changes over time, or assess quality of care. Interpreting this measure is not straightforward as it is affected by many factors, including disease stage (15), education (14), poverty index (14), urban/rural residence (9), marital status (14, 16), comorbidities (14, 15, 17), signs/symptoms at presentation (55), mammography history (14), hospital type (15–17, 47), whether diagnosis and treatment occur at the same or different hospitals (15), and year of diagnosis (15, 17). Treatment delay has been used to evaluate the quality of the NBCCEDP and the impact of policy changes (56). Although 94% of women meet their target of initiating treatment within 60 days of diagnosis, racial disparities exist and could be related to programmatic differences and geographic distinctions among other factors (7, 56). Therefore, interpreting and addressing this disparity is challenging even though participants share many characteristics and follow similar treatment guidelines.

Our findings also suggest that developing effective interventions for treatment delay require studies targeting specific populations. The impact of treatment delay on survival has been investigated in highly selected populations, including women in a program targeting underserved populations (22), with Medicaid coverage (54), with triple-negative breast cancer (57), receiving care at 2 hospitals served by the same providers following identical clinical protocols (47), and with metastatic disease (21). The results may not be generalizable, but these studies are helpful for identifying specific populations who may experience negative outcomes and the clinically relevant delay period. For instance, a study of Medicaid recipients in North Carolina found that a treatment delay of 60 days or more was associated with higher breast cancer–specific mortality among women with late-stage disease (hazard ratio, 1.85; 95% CI, 1.04–3.27), but not among women with early-stage disease (54). Characterizing and addressing the determinants of delay for this specific subgroup of Medicaid recipients could have a major public health impact.

One study limitation is that baseline interviews are conducted 5 months after diagnosis. Treatments subsequent to the first course of treatment may affect responses, and some factors may have changed during this interval. Recall bias may also influence the results. Detailed household composition information (e.g., number of children, wage earners, etc.) was not collected, which limits our ability to interpret findings for this determinant. We could not calculate poverty indices based on household size and income because we only collect information on income categories.

Several factors known to impact timeliness of care were not captured in our study. We did not have information on the health care facility where the women received care (e.g., urban/rural location, type, etc.). A greater number of comorbidities is associated with increased treatment delay (14, 15, 17) and African American race (14, 17), but was not assessed in our analysis. Finally, we did not collect information from participants about their interaction with providers. African American women are less trusting of their cancer treatment team (58). Providers also communicate differently with African American and White patients (59).

In conclusion, we found that the determinants of treatment delay vary by race, a finding that may help explain the conflicting literature on racial disparities. Further investigation is needed to determine the clinical relevance of the determinants we identified. Younger African American women may need additional support to ensure timely care comparable with White women. Our findings support targeting specific populations when identifying and addressing determinants of treatment delay.

No potential conflicts of interest were disclosed.

Conception and design: S.A. McGee, R.C. Millikan

Development of methodology: S.A. McGee, R.C. Millikan

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.A. McGee, R.C. Millikan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.A. McGee, D.D. Durham, C. Tse, R.C. Millikan

Writing, review, and/or revision of the manuscript: S.A. McGee, D.D. Durham

Study supervision: R.C. Millikan

The authors thank Susan Campbell for her expert guidance on the medical record data, the CBCS research team for their hard work to collect the data, and all of the study participants who shared their personal and medical information to make this research possible.

This research was funded in part by the University Cancer Research Fund of North Carolina and the National Cancer Institute Specialized Program of Research Excellence (SPORE) in Breast Cancer (NIH/NCI P50-CA58223).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.