Breast Cancer Presentation in an Urban Health Care Safety Net System

Abstract

We examined the Safety-Net referral process for breast diseases to identify factors contributing to delays within it. Each record was mapped to a time line beginning with first abnormality and concluding with definitive diagnosis/treatment. The median interval between first sign and definitive diagnosis/treatment was 93 days. Need for repeat imaging and missed visits prolonged the interval. System-and patient-specific factors were associated with delayed diagnosis/treatment in breast patients referred through the Safety-Net specialty clinic.

Introduction

Breast cancer is the most common non-cutaneous malignancy among women in the United States (U.S.), with approximately one in eight women (12.5%) developing the disease in the course of her lifetime.¹ Although the overall rate of breast cancer mortality is declining nationwide, socioeconomic and racial disparities in breast cancer morbidity and mortality not only persist in the United States but are, in fact, worsening.² In St. Louis, Missouri (which is more than 50% African-American³), approximately 40% of women diagnosed with breast cancer through the Safety Net’s specialty clinic – the system of referral by community health centers that provide primary care for more than 90% of the city’s un-and underinsured population⁴ – present with stage III and IV cancers. In contrast, at Siteman Cancer Center (the only National Cancer Institute (NCI)-designated Comprehensive Cancer Center within a 240-mile radius of St. Louis), to which most patients are referred via the private health care system, the majority of patients (85–90%) present with stage I or stage II disease (Lori Grove, Siteman Cancer Center Oncology Data Services, personal communication, 11 December 2006). Stage of disease at diagnosis has long been recognized as a significant prognostic factor: the five-year survival rate for stage I cancer is 88%, in contrast to 41–67% for stage III and 15% for stage IV.⁵ Thus, the observed disparities in breast cancer stage at diagnosis and subsequent outcomes between poor, medically underserved patients seen first at community health centers and insured patients seen first by private health care providers have raised concerns about the referral process for Safety-Net patients.^6,7

The breast cancer referral process of the St. Louis Safety-Net system often began with primary care evaluation at a community health center (CHC), followed by referral to a specialty surgical clinic, and concluding with definitive treatment at the Siteman Cancer Center (SCC). Once seen at the SCC, women referred through the Safety Net specialty clinic receive the standard of care afforded to similarly staged breast cancer patients with insurance who presented directly to the SCC. Because delays in presentation of symptomatic disease are associated with poorer survival,⁸ the high standard of care these women eventually receive does not mitigate the overall differences in morbidity and mortality observed between private health care and Safety-Net specialty clinic patients.⁹ In Missouri, relatively low breast cancer incidence (ranked 40^th in the nation) and a relatively high mortality rate (ranked 16^th in the nation) further support the likelihood that stage of disease at diagnosis contributes to high mortality rates.¹⁰

Delays in breast cancer diagnosis have been analyzed both as a “patient delay,” referring to a prolonged period between a patient’s discovery of breast cancer symptoms and medical evaluation, and as a “system delay,” referring to a prolonged period between medical evaluation and diagnosis and/or initiation of treatment.^11,12,13 The former has generally received more attention in breast cancer research than the latter, though study-to-study variability in how time intervals are defined has made it difficult to determine what constitutes timeliness versus delay when navigating the breast care pathway from detection of an abnormal breast sign or symptom to definitive treatment.^14,15 While the tumor doubling rate for a mammographically detectable tumor is approximately 260 days,¹⁶ no one suggests that an elapsed period of nearly nine months between detection and diagnosis is acceptable. Based on their clinical and research experience, Caplan et al. established that total intervals of 90 days – 30 days for the period from detection of a breast abnormality to diagnosis and 60 days for the period between diagnosis and treatment initiation – would represent an appropriate upper limit for timely care.¹⁷ Furthermore, in a systematic review of 87 studies, patients with delays of 3 months or more between onset of symptoms and initiation of treatment had 12% lower five-year survival versus those with a shorter interval.¹⁸

Quantifying the length of time it takes our city’s Safety-Net patients with breast disease to proceed through the breast care pathway was determined to be a critical first step in addressing this health disparity. The aim of this descriptive, hypothesis-generating study was to examine the Safety Net’s breast cancer referral process and to identify personal behaviors and institutional practices that might contribute to delays in breast cancer diagnosis and treatment in this population.

Materials and Methods

Following institutional review board (IRB) approval, we performed a retrospective chart review of 73 women consecutively seen at the Safety Net’s specialty surgical clinic between January 2005 and July 2007. Patients’ demographic data and visit information – including number of missed visits – were extracted from the medical record. Each patient’s record was mapped to a time line, with dates recorded for six “points of interest”:

1. First abnormal sign or symptom (identified by the patient or a health care provider),

2. First encounter with a CHC after abnormality noted,

3. Date of CHC referral to the specialty surgical clinic,

4. First patient contact with the specialty surgical clinic,

5. First appointment at the specialty surgical clinic, and

6. Definitive surgical treatment with confirmatory pathological diagnosis at the SCC or the conclusion that no further testing or treatment was needed.

The date on which pathologic diagnosis was obtained via surgical treatment was chosen as the final time point of interest because although a preoperative clinical diagnosis was given in most cases, pathological assessment was still required for definitive diagnosis, especially in cases of malignancy.

The five intervals between the six time points of interest were analyzed in the context of systems delays and missed patient visits. Patients missing dates for one or both of the relevant time points of interest were excluded; hence, most calculations were performed on sample sizes of fewer than 73 patients, our original total. A nonparametric rank-sum test was used to analyze each interval by race, presence of malignancy, and stage of malignant disease, if present. Fisher’s exact chi-square test was used to assess whether there were associations between race and a diagnosis of malignancy; race and stage of malignancy, if present; having a missing diagnosis and having missed one or more visits; having a missing diagnosis and having to have one’s mammogram(s) repeated or reinterpreted prior to referral to the SCC; having stage III or IV cancer and having missed one or more visits; or having stage III or IV cancer and having to have one’s mammograms repeated or reinterpreted. SAS 9.2 (Cary, NC) was used to perform all quantitative analyses.

Results

The sample of 73 women was 74% African-American; 45% were between the ages of 40–59 years of age, and 43% lacked health insurance (i.e., self-pay). Our sample represented a disproportionate number of women (77%) from four CHCs (See Table 1).

Table 1.

Distribution of Characteristics Within the Study Population

Characteristic	Study-wide n (%)	Study Population by CHC, n (%)
		Clinic A	Clinic B	Clinic C	Clinic D	Othera
Total number of women	73 (100)	3 (4)	21 (29)	24 (33)	8 (11)	17 (23)
Race
Caucasian	10 (13.7)	1 (33.3)	4 (19.0)	0 (0)	2 (25.0)	3 (17.6)
African American	54 (75.3)	2 (66.7)	12 (57.1)	23 (95.8)	4 (50)	13 (76.5)
Other	9 (12.3)	0 (0)	5 (23.8)	1 (4.2)	2 (25.0)	1 (5.9)
Health Core Coverage
Medicaid	10 (13.7)	0 (0)	2 (9.5)	5 (20.8)	2 (25.0)	1 (5.9)
Medicare	11 (15.1)	0 (0)	4 (19.0)	6 (25.0)	0 (0)	1 (5.9)
Self-pay	31 (42.5)	3 (100)	9 (42.9)	7 (29.2)	3 (37.5)	9 (52.9)
Indigent	11 (15.1)	0 (0)	2 (9.5)	3 (12.5)	3 (37.5)	3 (17.6)
Other	10 (13.7)	0 (0)	4 (19.0)	3 (12.5)	0 (0)	3 (17.6)
Diagnosis
Benign	23 (31.5)	2 (66.7)	6 (28.6)	8 (33.3)	3 (37.5)	4 (23.5)
Malignant	30 (41.1)	0 (0)	8 (38.1)	9 (37.5)	2 (25)	11 (64.7)
Missing	20 (27.4)	1 (33.3)	7 (33.3)	7 (29.2)	3 (37.5)	2 (11.8)

^aEncompasses all other CHCs at which study patients were seen

Twenty-three women (32%) had benign breast findings while 30 women (41%) were diagnosed with a form of breast malignancy. Among the women with cancer, 20 (67%) had ductal carcinoma in situ (DCIS), stage I, or stage II disease; 8 (27%) had stage III or IV cancer; and 2 (6%) had non-epithelial malignant disease that happened to be located in the breast (e.g., lymphoma, granular cell tumor). Twenty women (27%) had no documented clinical or pathological diagnosis or evidence of ever having received definitive treatment or discharge from clinic, though a few (<10%) of these women, based on physician and nursing notes, had signs and symptoms (including weight loss and palpable breast masses) that were suspicious for malignancy.

The number of days between first abnormal sign or symptom and definitive treatment ranged from 0 to 655, with a median of 93 days (n = 50). The median interval for time between first abnormality and first patient contact with the specialty surgical clinic was 36 days, with a range of 0 to 403 days (n = 61). Once at the specialty surgical clinic, women were seen and referred to the SCC relatively quickly, with a median interval of 12 days (range = 0–600, n = 62) between specialty surgical clinic appointment and definitive diagnosis and treatment at the SCC (See Table 2).

Table 2.

Median Intervals in Days Between Key “Dates of Interest”

		Median elapsed time (minimum - maximum) in days Sample size accounting for missing values
	Interval 1 First Abnormality to First CHC Encounter	Interval 2 First CHC Encounter to CHC Referral to Specialty Surgical Clinic	Interval 3 CHC Referral to Specialty Surgical Clinic to First Patient Contact with Specialty Surgical Clinic	Intervals 1–3 First Abnormality to First Patient Contact with Specialty Surgical Clinic	Interval 4 First Patient Contact with Specialty Surgical Clinic to Specialty Surgical Clinic Appointment	Intervals 5 Specialty Surgical Clinic Appointment to Definitive Diagnosis & Treatment	Intervals 1–5 First Abnormality to Definitive Diagnosis & Treatment
Study n=73	0 (0 – 205) n = 43	9 (0 – 383) n=44	13 (0 – 86) n=40	36 (0 – 403) n = 61	0 (0 – 525) n = 69	12 (0 – 600) n = 62	93 (0 – 655) n = 50

Relationships between race, interval length, and malignancy presence and stage were examined. There were no statistically significant differences between blacks and whites with regards to overall or subinterval lengths, presence of malignancy, or stage of malignancy, nor was there any statistically significant difference in interval length between different stages of cancer. However, the presence of malignancy correlated with statistically significant differences in intervals 3 (between the CHC referral to the specialty surgical clinic and first patient contact with the specialty surgical clinic; p=0.0079), 4 (between first patient contact with the specialty surgical clinic and the patient’s specialty surgical clinic appointment; p=0.0042), and 5 (between the patient’s specialty surgical clinic appointment and definitive diagnosis & treatment; p=0.0174) (See Table 2 for additional interval information). Specifically, intervals 4 and 5 were longest while interval 3 was shortest for women with cancer diagnoses. The difference in overall median interval length was longest for those with malignancy compared to those without malignancy or with no diagnosis, though it did not reach statistical significance (p=0.0648).

Thirty-two women (43.8%) had mammograms repeated or reinterpreted prior to definitive treatment and diagnosis; in this group, the median interval between first abnormality and definitive diagnosis/treatment was 136 days (range = 13–655 days, n = 22) versus 81 days (range = 0–558 days, n = 28) for those with adequate initial mammograms (p = 0.118).

Twenty women (27.4%) missed one or more visits, with a median interval of 110 days (range = 59–655 days, n = 10) versus 92 days (range = 7–434 days, n = 39) for those who missed no visits (p = 0.189). Women who missed one or more visits were significantly more likely to never receive definitive diagnosis and treatment (p = 0.0003).

Discussion

At thirty-six (36) days, the median interval in our cohort between a patient’s first abnormality and first contact with the surgical specialty clinic was longer than Caplan’s suggested standard of 30 days between detection of a breast abnormality and diagnosis of breast disease.¹⁹ This time period included three important sub-intervals: first, between the patient’s initial awareness of a breast abnormality and her first contacting her primary care provider (PCP) at a CHC (median = 0 days, n = 43); second, between the patient’s having been seen at a CHC and the patient’s being referred to the specialty surgical clinic (median = 9 days, n = 44); and third, between the patient’s being referred to the specialty surgical clinic and the patient’s being contacted by the specialty surgical clinic (median = 13 days, n = 40). Most women for whom this data was available (34 out of 43, i.e., 79%) reported immediately contacting their PCPs upon first knowing of a breast abnormality, indicating a high level of patient awareness about concerning breast signs and symptoms and the importance of seeking medical attention when such findings are discovered. Thus, given the relatively high level of patient awareness and activity with regards to promptly seeking medical care, potential sources of delay during this period might be ascribed to the process through which patients are scheduled for appointments at the CHCs and/or the process through which patients are referred by the CHCs to the specialty surgical clinic.

The median overall interval (from abnormality to definitive diagnosis/treatment) in our study was 93 days, which is greater than the 90-day standard recommended by Richards et al.²⁰ and by Caplan et al., whose median interval for a similar period in her 2000 AJPH publication was 48 days, with only 23% of patients having an overall interval length of more than 90 days.²¹ However, the median overall interval within our cohort, although longer than ideal, was less concerning than the range: 54% (27 out of 50) of the patients for whom this interval was calculated had median intervals that were longer than 90 days (ranging from 92 to 655 days), indicating that the process through which Safety-Net breast patients are ultimately evaluated and treated is not only fraught with delays but also idiosyncratic. Furthermore, for 2 of the 5 time sub-intervals examined, patients who were ultimately diagnosed with breast cancer were more likely to have had a longer delay than those who did not have cancer or who never received a diagnosis, and these findings were statistically significant.

Though this comparison was not statistically significant, women who had to have their mammograms repeated or reinterpreted had a longer median interval between their first being aware of a breast abnormality and ultimately receiving definitive diagnosis and treatment than women with adequate initial mammograms: 136 versus 81 days. City-wide variance in availability and quality of mammographic ser vices and record management is presumably reflective of the differences between CHCs and their catchment areas with regards to the capacity of their radiographic image storage facilities, the frequency of mammography van visits, and the availability of either internally or externally contracted radiologists to review scans. Precise information as to the radiographic resources of the different CHC branches is incomplete.

In the absence of real-time collection of data points that would assist in making the distinction between provider and patient delays, our record review included notation of any missed visits during the study period as a proxy for cases where delays might have been the result of patient unresponsiveness or inaccessibility; logistical or financial barriers, such as inconsistent access to transportation; or emotional barriers, such as fear. Although this comparison also failed to achieve statistical significance, women who missed one or more visits had a longer median interval between first breast abnormality and receiving definitive diagnosis and treatment than women who did not miss any visits: 110 versus 81 days. Notably, women who missed one or more visits were more likely to never receive definitive diagnosis and treatment, and this correlation was statistically significant (p = 0.0003). This finding suggests that missing diagnoses in our study sample were not missing at random, i.e., women who never received diagnoses or treatment differed significantly from those who did, at least with regards to their having missed one or more visits but possibly along other dimensions as well. Based on their medical records, it is not clear what happened to these women. This sub-analysis highlights the importance of aggressively following up with patients who miss even one scheduled appointment, and provides an argument for patient navigation practices to assist such at-risk patients.

In its regional, annually published Access to Care data reports, the St. Louis Regional Health Commission has not reported patient volumes by race for Safety-Net providers in the region (http://www.stlrhc.org, accessed April 26, 2011). However, nearly 75% of our study patients were African-American while only a little more than 50% of St. Louis City residents are African-American,²² indicating that a significant proportion of African-Americans utilize the Safety-Net ser vices compared to whites, as well as a potential higher level of under-and uninsured status amongst African-Americans in the region. The age-adjusted breast cancer mortality rate in Missouri is amongst the highest in the nation at 25.8 deaths per 100,000, with 35.3 breast cancer deaths per 100,000 among African-American women versus 25 breast cancer deaths per 100,000 among white women. Excess and racially disparate mortality rates at the state level are mirrored in local breast cancer mortality rates for all races and for African-American women living in St. Louis City (26 and 30.6 deaths per 100,000, respectively). A widening racial disparity is evident when mortality rates are compared: while rates for white women have steadily decreased over time, rates for African-American women with breast cancer have experienced a more gradual and less remarkable decline.²³ Stage and tumor pathology likely factor into the disparity in survival given the evidence that more aggressive disease, such as triple negative breast cancer, is more prevalent among African-American women.²⁴ Though the mechanisms behind this relationship remain elusive, African-American race is consistently associated with delays in breast cancer diagnosis and treatment.²⁵

Conclusion

There was quantitative evidence of delays in care amongst patients with breast findings referred through the Safety-Net specialty clinic. Factors associated with both the Safety-Net system (e.g., possible administrative delays in FQHC appointment scheduling and referral processes, variability in mammogram availability and quality) and patients (e.g., presence of malignancy, missing one or more scheduled appointments, and African-American race) may be associated with these delays in diagnosis and receipt of definitive treatment of breast abnormalities. Importantly, never receiving a definitive diagnosis or treatment correlated significantly with missing one or more clinic appointments, highlighting the ongoing importance of patient education about the importance of keeping one’s scheduled clinic appointments and the need for aggressive administrative follow-up of vulnerable patients at risk for falling through the “holes” in the Safety Net.

Limitations associated with retrospective chart reviews restricted our ability to collect and/or consider data on specific patient factors that may have contributed to delays. Follow-up inter views with patients were not conducted in light of the potential for recall bias. We were unable to confirm with certainty the extent to which delays were attributable to patient-specific circumstances such as accessibility by phone, availability of transportation to appointments, or social and environmental factors that might impact compliance.

As the preliminary findings of the current study seemed to support our initial concerns about delays in the Safety-Net referral process, we shifted our attention from collecting data for this retrospective analysis (with its associated design and analytic biases and limitations) to designing and funding a prospective study that would provide more information about potential patient and/or systemic factors contributing to delays in diagnosis and definitive treatment. This preliminary review provided data support for the prospective study we are currently conducting, in which we compare the experiences of Safety-Net and private healthcare patients presenting with breast symptoms or abnormalities to the SCC. By documenting the experiences of Safety-Net patients with breast disease, we hope to obtain a better understanding of what factors – at the level of both individuals and institutions – are perpetuating the observed inequities in breast cancer care in St. Louis. We anticipate both the study design and analysis of the prospective project will address many of the shortcomings of this retrospective study.

Biography

Julie A. Margenthaler, MD, (above left), MSMA member since 1998, Oluwadamilola M. Fayanju, MD, (above right) Jennifer R. Tappenden, RHIA, Courtney E. Beers, MPH, and Bettina F. Drake, PhD, are in the Department of Surgery; Donna B. Jeffe, PhD, is in the Department of Internal Medicine, Division of Health Behavior Research; Feng Gao, PhD, is in the Division of Biostatistics; all are at the Washington University School of Medicine in St. Louis. Both Drs. Margenthaler and Jeffe also practice at the Siteman Cancer Center at Barnes-Jewish Hospital in St. Louis.

Contact: margenthalerj@wudosis.wustl.edu