Physician Referral Patterns Impact Survival of Older Women with Metastatic Breast Cancer

Special Article - Geriatric Medicine

Gerontol Geriatr Res. 2017; 3(2): 1033.

Physician Referral Patterns Impact Survival of Older Women with Metastatic Breast Cancer

Tkaczuk KHR¹*, Yong C¹, Onukwugha E¹, Mullins CD¹ and Hussain A¹,²

¹University of Maryland School of Medicine, USA

²Veterans Affairs Medical Center, USA

*Corresponding author: Katherine HR Tkaczuk, University of Maryland School of Medicine, Professor of Medicine, Director, Breast Evaluation and Treatment Program, University of Maryland Greenebaum Comprehensive Cancer Center, 22 South Greene St, Baltimore, USA

Received: August 09, 2017; Accepted: November 22, 2017; Published: November 29, 2017


Objective: Contact with Primary Care Physicians (PCP) has been found to be associated with improved outcomes, including decreased mortality, among women with breast cancer. The impact of the medical oncologist in the context of PCP contact on outcomes of women with metastatic BC (mBC) is not well defined. We examined if PCP contact prior to mBC diagnosis affects survival independently of or is in some measure dependent on subsequent medical oncologist contact.

Methods: This analysis used linked SEER (Surveillance, Epidemiology, and End Results) and Medicare data on women aged 65+ with incident mBC diagnosed during 2007-2009. Using Cox proportional hazards models, we examined the influence of PCP contact within the one-year period preceding mBC diagnosis on the probability of post-diagnosis medical oncologist visit and on overall mortality, adjusting for baseline patient characteristics.

Results: Of 2,066 women (mean age 77 years), 728 (35.2%) did not have PCP contact within one year before mBC diagnosis. Women with three or more PCP visits were more likely to see a medical oncologist (3-4 visits vs. 0 visits: HR=1.29; 95% CI=1.11-1.50; 5+ visits vs 0 visits: HR=u1.41; 95% CI=1.22- 1.63). Older age, African American race, and lower socioeconomic status were statistically significantly associated with decreased probability of medical oncologist visit. All-cause mortality was lower among those with post-diagnosis medical oncologist contact (53.5% - 55.8%), and higher (86.7% - 89.8%) among those without medical oncologist contact (p<0.01).

Conclusion: Pre-diagnosis primary care contact impacts positively referral to medical oncology subspecialty care while the post-diagnosis access to medical oncology care contributes significantly to lower all-cause mortality in older patients with incident mBC.

Keywords: Metastatic breast cancer; Primary care; Physician specialist; Medical oncology


AJCC-TNM: American Joint Committee on Cancer Tumor- Node-Metastasis staging; AMA: American Medical Association; CCI: Charlson Comorbidity Index; CI: Confidence Interval; ER: Estrogen Receptor; HR: Hazard Ratio; IQR: Interquartile Range; mBC: Metastatic Breast Cancer; PCP: Primary Care Physician; PR: Progesterone Receptor; SD: Standard Deviation; SEER: Surveillance, Epidemiology, and End Results Cancer Registry


Compared to early stage breast cancer, metastatic Breast Cancer (mBC) is associated with poorer prognosis with a median overall survival of 2 to 3 years [1-3]. Although mBC is not considered curable, it is generally treatable with the primary goals of care being to optimize survival time and quality of life [4]. Comprehensive care for women with breast cancer generally includes evaluation and management by a multidisciplinary team, including medical oncologists, radiation oncologists, breast surgeons as well as Primary Care Physicians (PCP) [5,6]. Barriers or delays to cancer specialist referral following diagnosis of cancer could have negative implications on patient satisfaction with care [7], treatment receipt [8,9], and clinical outcomes such as overall survival [10].

Although patient referral to a medical oncologist for cancer treatment evaluation is considered a key step in the pathway to treatment receipt [8,9,11], there is limited published evidence on the factors that influence referral to medical oncologists, particularly in the advanced breast cancer setting. The potential barriers to patient referral to specialty oncology care include patient-level factors such as socioeconomic constraints, and system-level factors such as restricted provider networks and preauthorization requirements [12]. In addition to these factors, an individual’s ability to engage in the healthcare system during the pre-diagnosis period, including regular contact with a PCP, could also influence referral to specialty care following their cancer diagnosis.

Previous studies have reported that PCPs play an important role in the early detection of breast cancer through cancer screening [13,14]. Increasing number of PCP visits in the period prior to BC diagnosis has been found to be associated with improved BC-related outcomes, including lower odds of late-stage diagnosis, and lower BC-specific and overall mortality [15,16]. Specifically, among 90,537 Medicare women with BC diagnosed during the years 1994-2005, women who had 10 or more PCP visits in the 24-month period prior to BC diagnosis had 41% lower BC mortality, and 27% lower overall mortality, compared with women who had 0 to 1 PCP visit [15]. However, it is uncertain if primary care in the pre-diagnosis period influences mortality independently, or through the effect of primary care on referral to medical oncology. Therefore, we undertook the present study to address the question of whether PCP contact with in one year prior to incident (newly diagnosed) mBC affects BC outcomes independently or is in some measure dependent on subsequent medical oncologist contact. The potential role of the PCP in this context would be particularly relevant to the elderly women population since PCPs are likely to play an increasingly important role in the healthcare of aging women, and the management of patients with cancer [17].

The study objective was to examine the association between prediagnosis PCP contact and post-diagnosis medical oncologist visit among older women with newly diagnosed mBC, and the influence of physician contact on overall survival. We hypothesized that women with primary care contact in the pre-diagnosis period were more likely to visit a medical oncologist after mBC diagnosis. We also explored whether the intensity of primary care contact (number of pre-diagnosis PCP visits) had any influence on the probability of post-diagnosis medical oncologist visit.

Patients and Methods

Study design and study population

This was a retrospective cohort analysis of linked Surveillance, Epidemiology, and End Results cancer registry and Medicare claims data (SEER-Medicare) on female Medicare beneficiaries with incident mBC diagnosed during January 2007 to December 2009. The study was approved by the University of Maryland Institutional Review Board. Stage of breast cancer was determined from SEER in accordance with the American Joint Committee on Cancer Tumor-Node-Metastasis (AJCC-TNM) staging, 6th edition [18]. Patients were included in the study sample if they were age 66 or older at the time of diagnosis, had continuous enrollment in Medicare Parts A and B in the 12 months prior to diagnosis month, and survived for at least 30 days after diagnosis. Patients were excluded if they had any of the following: 1) history of any cancer (excluding non-melanoma skin cancer) within 5 years prior to the BC diagnosis; 2) unknown diagnosis month or year; or 3) incident post-mortem BC diagnosis. Medicare claims data from 2006 to 2011 were used to capture information on physician visits during the pre-diagnosis and post-diagnosis periods.


The primary outcomes were post-diagnosis medical oncologist visit, and overall mortality following diagnosis. The independent variable of interest was a binary indicator for any PCP visits during the one year pre-diagnosis period (0 PCP visits vs. 1 or more visits). To explore if the intensity of primary care contact affected referral to a medical oncologist, PCP visits in the pre-diagnosis period was categorized as follows: 0 PCP visits vs. 1-2 visits vs. 3-4 visits vs. 5 or more visits.

Medicare claims from the Carrier Claims (National Claims History) and Outpatient files were used to capture physician visits in the ambulatory setting. Physician specialty was determined from the American Medical Association (AMA) Physician Masterfile. Medical oncologists were identified from AMA specialty codes for hematology and/or oncology. PCPs were identified from specialty codes for general practice, family practice, internal medicine, or geriatric medicine. Radiation oncologists and surgical oncologists were identified based on specialty codes for radiation oncology and surgical oncology, respectively.

Statistical analyses

Descriptive characteristics of the study sample were presented using frequency distributions for categorical variables and median and inter quartile values for continuous variables. Chi-square tests were used to determine the bivariate associations between patient characteristics and any pre-diagnosis PCP visits. Cox proportional hazards models were estimated to examine the covariate-adjusted association between pre-diagnosis PCP contact and the probability of post-diagnosis medical oncologist visit. We also estimated separate Cox proportional hazards models for all-cause mortality in the full sample and among patients with at least 1 medical oncologist visit to determine if pre-diagnosis PCP contact had any additional effect on overall survival, conditional on a medical oncologist visit.

The following potential confounding demographic, clinical, and contextual variables were included in the regression models: age group at diagnosis, race/ethnicity, Estrogen Receptor/Progesterone Receptor (ER/PR) status at diagnosis, tumor differentiation (poorly or undifferentiated tumor), comorbidity burden at baseline as measured by the Charlson Comorbidity Index (CCI), screening and preventive care services (including screening/diagnostic mammography, and flu vaccination) in the year prior to diagnosis, a single proxy measure for poor performance status (any use of wheelchair, walking aid, oxygen, skilled nursing facility service, or hospitalization in the year prior to diagnosis), diagnosis year, and census region of SEER registry (Northeast, West, Midwest, or South). The regression models also included a proxy measure for low income i.e., an indicator for any state buy-in in the year prior to diagnosis. Medicare buy-in benefits are generally operated by state Medicaid programs, and are provided to low-income Medicare beneficiaries to cover their Medicare premiums, deductibles, and copayments [19].

Interaction terms between pre-PCP visits and patient factors including age, race, and state buy-in were included in the Cox proportional hazard models to test for the presence of statistically significant interactions. The final regression model did not include any interaction terms as there were no statistically significant interactions identified. The proportional hazard assumption with respect to the pre-PCP visits variable was also tested. Time-invariant adjusted Hazard Ratios (HR) was reported as the proportional hazard assumption was not violated. All statistical tests were two-tailed with a 0.05 cut-off value for statistical significance. All statistical analysis was conducted using Version 9.3 of the SAS System.


Descriptive characteristics of study cohort

The study cohort included 2,066 female Medicare recipients with incident metastatic breast cancer diagnosed during 2007 to 2009. The mean age of the study sample was 77 years, and the sample comprised 81% who were non-Hispanic white and 11% who were non-Hispanic African American. Of the 2,066 women, 728 (35.2%) did not have any PCP visit during the one year prior to diagnosis. Among the 1,338 women with at least 1 PCP visit in the year prior to diagnosis, the average number of PCP visits was 5 visits (SD=3.8; Inter Quartile Range (IQR) = 2-6 visits). Descriptive characteristics of the sample, stratified by any PCP visits in the one year prior to diagnosis, are shown in Table 1. Overall, compared to those with at least 1 PCP visit, patients with no PCP visits in the pre-diagnosis period were younger, had lower comorbidity burden as measured by the Charlson Comorbidity Index (CCI), and had a lower proportion with poor performance status (any use of wheelchair, walking aid, oxygen, skilled nursing facility service, or hospitalization in the year prior to diagnosis) (p<0.01 for all).