Treatment of local breast carcinoma in Florida†
The role of the distance to radiation therapy facilities
The study protocol was approved by the University of Miami and the University of Florida IRBs and by the Florida Department of Health.
Breast-conserving surgery combined with radiation (BCSR) is the recommended alternative treatment to mastectomy for local breast carcinoma. However, limited access to healthcare may result in more extensive surgical treatment. The effect of distance to radiation therapy facilities on the likelihood of receiving BCSR was examined in Florida.
Local breast carcinomas reported to Florida's statewide registry between July, 1997, and December, 2000 were linked to the Agency of Healthcare Administration inpatient and outpatient databases to supplement the registry's treatment data, resulting in 18,903 local breast carcinoma cases treated with BCSR or mastectomy. The odds of receiving BCSR were modeled as a function of distance to the closest radiation therapy facility, adjusting for health insurance, age, race/ethnicity, and marital status.
Distance to the closest radiation therapy facility was negatively associated with BCSR, with the odds ratio (OR) decreasing by 3% per 5-mile increase in distance. Compared with the uninsured, privately insured women were 49% more likely to receive BCSR (OR of 1.49; 95% confidence interval [95% CI], 1.20–1.86) and Medicare patients were 37% more likely (OR of 1.37; 95% CI, 1.09–1.72). Age at diagnosis was negatively associated, reducing the odds of BCSR by 1% per year increase in age. Compared with white non-Hispanic, Hispanic women were 38% less likely to receive BCSR (OR of 0.62; 95% CI, 0.55–0.71). Married women were 23% more likely to receive BCSR compared with singles (OR of 1.23; 95% CI, 1.09–1.40); women who were separated, divorced, or widowed did not differ significantly from single women.
Distance to radiation therapy facilities may negatively impact the likelihood of BCSR in Florida. Age at diagnosis, insurance type, race/ethnicity, and marital status were associated with BCSR. Future efforts should target the uninsured, Hispanics, the elderly, and the unmarried women to reduce disparities in the administration of BCSR for local breast carcinoma. Cancer 2006. © 2005 American Cancer Society.
The treatment of early stage breast carcinoma has changed drastically in the past 20 years. During this period, clinical trials indicated that breast-conserving surgery with radiation therapy (BCSR) produced similar results with regard to rates of second primary cancers, contralateral breast carcinomas, or metastases and resulted in similar long-term survival compared with mastectomy. As a result, multiple medical groups have recommended that women be given a choice regarding therapy of early-stage breast carcinoma, particularly BCSR.1–5
Despite clinical guidelines, treatment patterns have not reflected adherence to these recommendations. Multiple studies indicated that the elderly, race/ethnic minorities, the poor, and the uninsured are less likely to receive standard therapy.6–12 It has been reported, for example, that after breast-conserving surgery for local stage breast carcinoma, black non-Hispanic women receive radiation therapy less frequently compared with white non-Hispanic women.10, 11 Cancer treatment has been shown to vary by health insurance as well. Several studies found that the uninsured and women with Medicaid were less likely to receive radiation therapy after breast-conserving therapy for local stage breast carcinoma compared with women with private insurance or Medicare.9, 13 Conversely, a single-institution study comparing the medically indigent to the insured patients found no difference in the use of breast conservation therapy.14
Marital status, acting as a proxy for social support, has been shown to influence the receipt of medical care and ultimately impacts screening, treatment, and the survival of cancer patients.15–18 Social support may be especially important for treatment modalities such as radiation that involve the logistics of multiple visits to treating facilities in addition to dealing with radiation side effects.
Finally, geographic variations have been observed in the treatment of local stage breast carcinoma,11, 19–22 raising the question of whether women are less likely to receive BCSR as the distance from their place of residence to radiation therapy facilities increases.
In this population-based study, we linked Florida incident cancer registry data to administrative inpatient and outpatient data to compile a more complete treatment profile and test the hypothesis that the further away the patients lived from radiation therapy facilities, the higher was the likelihood of having mastectomy rather than BCSR. In our analysis, we accounted for the impact of other significant factors reported by researchers, including health insurance, age at diagnosis, race-ethnicity, and marital status.
MATERIALS AND METHODS
Data Sources and Linkage
The Florida Cancer Data System (FCDS) is the statewide population-based incident registry and is a member of the North American Association of Central Cancer Registries (NAACR), collecting all newly diagnosed primary cancers in Florida since 1981. Audits conducted by the NAACCR have estimated case reporting to be 99.4% complete.23 The Florida Agency for Health Care Administration (AHCA) maintains two databases (Hospital Patient Discharge Data [HPDD] and Ambulatory Outpatient Data [AOD]) on all patient encounters within hospitals and freestanding ambulatory surgical and radiation therapy centers.24 All licensed facilities in Florida have been required to report all discharges and outpatient encounters to AHCA since 1988 and 1997, respectively. The AHCA datasets used in this study contained diagnoses and procedures performed during every hospitalization or outpatient encounter in the state of Florida for the period 1997–2000.
AHCA collects demographic information and personal identifiers. The FCDS records were linked to the AHCA records. The linkage was performed at the patient level, using a probabilistic algorithm based on social security number, date of birth, gender, race, and county of residence at the time of diagnosis. Approximately 94% of the local breast carcinoma records were linked to either an inpatient or an ambulatory AHCA record via this procedure.
We used the census-provided Topological Integrated Geographic Endocoding and Reference files (TIGER)25 for geocoding the patients' addresses at diagnosis and assigned longitude and latitude. Approximately 98% of the patients' addresses were geocoded at the street level. Using Geographic Information Systems (GIS) software, we identified the closest radiation therapy facility to the patient and computed the Euclidean (shortest) distance to that facility in miles.
Study Population and Data
The initial population for this study originated from the 26,423 primary breast carcinomas diagnosed at SEER local stage26 in female Florida residents between July 1997 and December 2000. FCDS collects all components of first course of treatment including: site-specific surgery, radiation therapy, chemotherapy, and endocrine therapy administered in the first 4–12 months of a breast carcinoma diagnosis.27
Any missing or uninformative treatment information in the FCDS records was supplemented by the corresponding AHCA information, via linkage of the two datasets. Through this process, 13,331 (50%) were linked to AHCA inpatient records and 21,868 (83%) were linked to AHCA outpatient records, with an overall linkage success rate of 94%. After the enhancement process, 447 cases had missing treatment information, 211 had no surgery, 3047 had lumpectomy but no radiation, 12,569 had lumpectomy with radiation, 8219 had mastectomy without radiation, and 1930 had mastectomy with radiation. Because the purpose of this study was to test the hypothesis that the longer distance to radiation facilities may have resulted in more extensive surgical treatment, we compared the cases that received the two guideline-based treatment modalities for which the majority of the patients had an option, namely, the cases that received mastectomy (only) to those receiving BCSR. Elimination of records with missing information in the treatment, geocoding, or other analysis variables reduced the study population to 11,354 cases treated with BCSR and 7549 cases treated with mastectomy.
The FCDS surgery and radiation therapy data were used for this analysis when available; if they were unavailable (missing or coded as no treatment administered) or uninformative (nonspecified surgery), the corresponding linked AHCA inpatient records were examined for surgery and radiation data separately. If the inpatient record was unavailable or uninformative, then the outpatient record was examined for treatment information. All procedure fields of the two AHCA datasets were used in this process. Two enhanced surgery and radiation therapy variables were created, containing the most complete breast carcinoma treatment administered to the patient. For this analysis, we compared two treatment choices: BCSR and mastectomy (only). All forms of radiation therapy were combined into a dichotomous variable (i.e., yes/no radiation). Breast-conserving surgery included the range of procedures encompassing less than a full mastectomy. The mastectomy category included: simple, modified radical, radical, and extended radical mastectomy.
Age at diagnosis and distance to radiation therapy facilities were used as continuous variables in the analysis. Race and ethnicity were combined into one variable containing the mutually exclusive categories: white non-Hispanic, black non-Hispanic, or Hispanic. Marital status was classified as: single, married, separated or divorced, or widowed. Insurance at the time of diagnosis was grouped into: uninsured, private, Medicare, Medicaid.
After the univariate analyses, a multivariate logistic regression model was fitted to estimate the odds of receiving BCSR compared with mastectomy. The covariates were: age at diagnosis, race/ethnicity, marital status, health insurance type, and distance to the closest radiation therapy facility. Interaction terms were also tested for selecting the model that best fitted the data. For the geographic component of the analysis we used ArcView 3.228 and for statistical analysis we used SAS v8.29
The study population consisted of 18,903 cases: 7549 cases that received mastectomy and 11,354 cases that received BCSR (Table 1). The average age at diagnosis was 64.9 years; 87.2% of the tumors were diagnosed in white, non-Hispanics; 7.5% in Hispanics; and 5.3% in black non-Hispanics. Approximately 59% of the women were married, 22.5% were widows, 9.6% were separated or divorced, and 8.4% were single. At diagnosis, 47.6% had private insurance, 48.4% had Medicare, 2.4% were uninsured, and 1.6% had Medicaid. The median Euclidean distance from the patients' residence at diagnosis to the closest radiation therapy facility was 5.9 miles (mean, 9.0, standard deviation [SD] = 9.0).
Table 1. Characteristics of Patients with Local Breast Carcinoma, by Type of Treatment, Florida, July 1, 1997 to December 31, 2000
|Age at diagnosis in yrs|| || || |
| Mean (SD)||63.9 (12.8)||66.5 (13.7)||65.2 (13.3)|
| Distance to closest radiation facility in miles|| || || |
| Mean|| 8.8 (8.8)|| 9.3 (9.2)|| 9.0 (9.0)|
|Race/ethnicity|| || || || || || |
| White, non-Hispanic||10,064||88.6||6420||85.0||16,484||87.2|
| Black, non-Hispanic||585||5.2||417||5.5||1002||5.3|
|Marital status|| || || || || || |
|Insurance|| || || || || || |
There was variability in the treatment of local breast carcinomas among various subgroups. Women who received BCSR were younger, with a median age of 66 years at the time of diagnosis (vs, 69 yrs in the mastectomy group), and lived closer to radiation treatment facilities (median of 5.8 miles) compared with those that received mastectomy (median of 6.1 miles). Approximately 49.8% of the local stage cancers diagnosed among Hispanics were treated with BCSR, compared with 58.4% in the black non-Hispanics and 61% in the white non-Hispanics (Table 2). Local breast carcinomas diagnosed among widows or single women were less often treated with BCSR (52.4% and 58.2%, respectively) compared with those diagnosed among married and separated or divorced women (61.6% and 63%, respectively). Insurance played an important role as well, with women insured by Medicaid being the least likely to receive BCSR (51.9%), followed by the uninsured (53.3%) and those with Medicare (56.5%); the privately insured were the most likely to receive BCSR (64.3%).
Table 2. Frequency of BCSR and OR Estimates for Local Breast Carcinomas Treated with BCSR versus Mastectomy Florida, July 1, 1997 through December 31, 2000 (N = 18,903)
|Age at diagnosis|| || || || |
| Per 1-yr increase|| || ||0.99||(0.98-0.99)|
| Per 5-yr increase|| || ||0.94||(0.92-0.95)|
| Per 10-yr increase|| || ||0.88||(0.85-0.91)|
|Distance from closest radiation facility|| || || || |
| Per 5-mile increase|| || ||0.97||(0.95-0.99)|
| Per 10-mile increase|| || ||0.94||(0.90-0.98)|
| Per 15-mile increase|| || ||0.91||(0.86-0.96)|
| Per 20-mile increase|| || ||0.88||(0.82-0.95)|
|Race/ethnicity|| || || || |
| White, non-Hispanic||16,484||61.1||1|| |
| Black, non-Hispanic||1002||58.4||0.88||(0.75-1.02)|
|Marital status|| || || || |
| Single||1579||58.2||1|| |
|Insurance type|| || || || |
| Uninsured||462||53.3||1|| |
To further explore the findings and the combined effect of all these factors on the odds of receiving BCSR versus mastectomy, we tested various logistic regression models. The model that best fit the data included: age at diagnosis, race/ethnicity, facility type, marital status, insurance type, and distance to the closest radiation therapy facility (Table 2). Various interaction terms were tested, but none were statistically significant.
With regard to our hypothesis, the distance to the closest radiation therapy facility was negatively associated with BCSR, with the odds ratio (OR) of receiving BCSR decreasing by approximately 3% for every 5-mile increase in the distance to radiation treatment facilities. In addition, compared with the uninsured, the privately insured were 49% more likely to receive BCSR (OR of 1.49; 95% confidence interval [95% CI], 1.20–1.86) and the Medicare insured were 37% more likely (OR of 1.37; 95% CI, 1.09–1.72). The age at diagnosis was a predictor of BCSR for local breast carcinoma, reducing the odds by 1% per year of increase in age. Compared with white non-Hispanic women, Hispanics were 38% less likely (OR of 0.62; 95% CI, 0.55–0.71) to receive BCSR. Married women were 23% more likely to receive BCSR compared with single women (OR of 1.23; 95% CI, 1.09–1.40); women who were separated or divorced and widows were similar to single women.
Access to healthcare is recognized as an important component influencing the overall population health. Affordability and accessibility are two dimensions of access to healthcare,30 with affordability relating to health insurance and accessibility to geographic access. In our study, we evaluated the joint effect of these two dimensions in the administration of BCSR compared with mastectomy for local breast carcinomas diagnosed in the state of Florida during the period July 1997, to December 2000. We used the distance from the residence at the time of diagnosis to the closest radiation therapy facility as a reasonable measure of geographic accessibility31 and the type of healthcare insurance at the time of diagnosis as a measure of affordability of healthcare.32
Travel distance to a radiation therapy facility may influence whether a patient receives postoperative breast irradiation, as it involves daily treatments for 5–6 consecutive weeks. In a New Mexico study of women with local breast carcinoma that had received breast-conserving therapy, Athas et al.22 found an inverse relationship between distance and subsequent use of radiation therapy. In the current study we compared two guideline-based treatment modalities for which the majority of patients have a choice between BCSR, which involves less extensive surgery and several subsequent visits to radiation therapy facilities, and mastectomy, which involves more extensive surgery but no visits to radiation therapy facilities. Our findings confirm that women who lived further from radiation therapy facilities were less likely to be treated with BCSR, after taking into account other known confounders, measurable by the registry data.22
Health insurance type has been confirmed as a factor leading to differential treatment by many studies, with the exception of a study conducted in Virginia,14 in which researchers found no difference in BCSR between medically indigent and insured patients with early-stage breast carcinoma who were treated at a particular academic institution that facilitated transportation. That study offers valuable lessons but it is most likely more indicative of the treatment administered in the particular institution. In the current study we found that women with private insurance or with Medicare were 37–49% more likely to receive BCSR for local breast carcinoma compared with the uninsured. In contrast, in our previous study of treatment of regional breast carcinoma in Florida33 we found that health insurance coverage impacted the odds of receiving systemic therapy differentially, depending on whether the reporting facility was an American Association of Medical College teaching facility, with the Medicare patients being the least likely to receive systemic therapy. These controversial findings raise questions regarding the differential treatment of Medicare-insured breast carcinoma patients, depending on the stage of the disease and the treatment modality. It is possible that predictors of surgery, radiation therapy, and systemic therapy are different, and more research is needed to elucidate the reasons for these differences.
Consistent with other researchers,34–36 we found that local stage breast carcinomas diagnosed in older women were more likely to be treated with mastectomy compared with those diagnosed in younger women. This finding could be attributed partly to comorbidity, but others have reported that age remains an independent determinant of treatment even after accounting for comorbidity.37 Married and separated or divorced women had higher odds of receiving BCSR compared with single women and widows. We tested for interactions between insurance type and marital status but none were significant, suggesting that the trend observed is not of a financial nature, but rather reflects the underlying issue of social support networks.
As in our previous study of treatment of regional breast carcinoma,33 we found differential treatment depending on the patient's race/ethnic origins. The same pattern was observed in the current study. In particular, Hispanic women were 38% less likely to be treated with BCSR compared with white, non-Hispanic women, whereas black non-Hispanic women were 12% less likely. This repeated pattern observed for Hispanics is not surprising. In a study of patients eligible for breast-conserving surgery (BCS) conducted in a managed care environment, Legorreta et al.38 reported that Hispanic women were 36% less likely to be treated with BCS compared with white women. Access to preventive care has been shown to be worse for Latinas compared to non-Latinas, but when accounting for socioeconomic level this difference was attenuated.39 Another report on health insurance trends indicated an increased access to healthcare gap and showed that Latinos are more disconnected overall from the healthcare system than whites.40 Another specific factor linked to the increased use of BCSR is access to multidisciplinary consultation, in particular, consultation with a radiation oncologist.41 For Hispanic women, the combination of socioeconomic and cultural barriers may decrease the access to healthcare; moreover, to special consultation regarding specific treatment options, which in turn may result in the lower use of BCSR.
Strengths and Limitations of the Study
In our study we used Euclidean distance to the closet radiation treatment facility because of its simplicity of calculation, disregarding traveling time or geographic barriers that may additionally impede access.42 Our intent was to demonstrate an association rather than quantify it. An actual distance on a road network and/or traveling time may be more appropriate for accurately quantifying the association found.
Of concern are potential discrepancies between the registry recorded treatment and the actual treatment administered to the patients. Validation studies are needed for assessing the concordance between the two.
In addition to distance to radiation therapy facilities, there are additional factors that may influence the receipt of mastectomy versus BCSR. Patient choice or medical contraindications such as pregnancy, connective tissue disease, previous irradiation of the breast, or large tumor size in proportion to breast size43 can make mastectomy the best choice for some women. These issues, however, cannot be assessed using registry or administrative data.
In patterns of care studies, the completeness and accuracy of the registry's treatment data has been questioned.9, 44–46 To overcome this problem we supplemented the cancer registry's database with information from an administrative database and created a more complete treatment profile. Through the AHCA datasets, specific surgery information was retrieved for 13% of local breast carcinoma records with uninformative surgery and for 35% of the cases that did not have radiation therapy reported according to the FCDS records. The overall improvement in the radiation therapy data reached 36% for all local breast carcinomas.
This population-based study describes treatment patterns for local stage breast carcinoma and offers potential explanations for the use of BCSR in Florida. We found that distance to radiation therapy facilities has an inverse relationship with BCSR use. In addition, we found that the uninsured and women with Medicaid were less likely to be treated with BCSR compared with Medicare and privately insured women, and raises questions regarding the differential treatment of Medicare patients depending on the stage of the disease and the treatment modality. Finally, this study identified a pattern among Hispanic women of being the least likely to receive BCSR, suggesting the need for targeting them more than any other race/ethnic group in Florida. Facilitating access to radiation therapy facilities and offering pretreatment counseling could potentially increase the receipt of BCSR among Hispanic women.
The authors thank the Florida abstractors for collecting the data and enabling studies such as this one and the Florida Cancer Data System and the Florida Department of Health for their support. They also thank the Minnesota Cancer Registry for providing the X-walk54 software, for the conversion of CPT codes to SEER treatment codes, and the Agency of Health Care Administration for allowing us to use their data for this study. They thank Mike Jeffe for assistance in the GIS component of this study.