- © 2013 by American Society of Clinical Oncology
Barriers to Breast and Colorectal Cancer Survivorship Care: Perceptions of Primary Care Physicians and Medical Oncologists in the United States
- Katherine S. Virgo, Emory University, Atlanta, GA; Catherine C. Lerro, Yale University, New Haven, CT; Carrie N. Klabunde, Division of Cancer Control and Population Sciences, National Cancer Institute, Bethesda, MD; Patricia A. Ganz, Jonsson Comprehensive Cancer Center, University of California Los Angeles, Los Angeles, CA; and Craig Earle, Institute for Clinical Evaluative Sciences and University of Toronto, Toronto, Ontario, Canada.
- Corresponding author: Katherine S. Virgo, PhD, MBA, Department of Health Policy and Management, RSPH, Emory University, 1360 Stephens Dr NE, Atlanta, GA 30329; e-mail: kvirgo{at}emory.edu.
-
Presented at the 47th Annual Meeting of the American Society of Clinical Oncology, Chicago, IL, June 3-7, 2011 (physician-related barriers) and the Annual Research Meeting of AcademyHealth, Seattle, WA, June 12-14, 2011 (patient-related barriers).
Abstract
Purpose High-quality, well-coordinated cancer survivorship care is needed yet barriers remain owing to fragmentation in the United States health care system. This article is a nationwide survey of barriers perceived by primary care physicians (PCPs) and medical oncologists (MOs) regarding breast and colorectal cancer survivorship care beyond 5 years after treatment.
Methods The Survey of Physician Attitudes Regarding the Care of Cancer Survivors was mailed out in 2009 to a nationally-representative sample (n = 3,596) of US PCPs and MOs. Ten physician-perceived cancer survivorship care barriers/concerns were compared between the two provider types. Using weighted multinomial logistic regression, we modeled each barrier, adjusting for physician demographics, reimbursement, training, and practice characteristics.
Results We received responses from 2,202 physicians (1,072 PCPs; 1,130 MOs; 65.1% cooperation rate). In adjusted patient-related barriers models, MOs were more likely than PCPs to report patient language barriers (odds ratio, [OR], 1.72; 95% CI, 1.22 to 2.42), insurance restrictions impeding test/treatment use (OR, 1.42; 95% CI, 1.03 to 1.96), and patients requesting more aggressive testing (OR, 4.08; 95% CI, 2.73 to 6.10). In adjusted physician-related barriers models, PCPs were more likely to report inadequate training (OR, 3.06; 95% CI, 2.03 to 4.61) and ordering additional tests/treatments because of malpractice concerns (OR, 1.87; 95% CI, 1.20 to 2.93). MOs were more likely to report uncertainty regarding general preventive care responsibility (often/always: OR, 1.97; 95% CI, 1.13 to 3.43; sometimes: OR, 2.16; 95% CI, 1.60 to 2.93).
Conclusion MOs and PCPs perceive different cancer follow-up care barriers/concerns to be problematic. Resolving inadequate training, malpractice-driven test ordering, and preventive-care responsibility concerns may require continuing education, explicit guidelines, and survivorship care plans. Reviewing care plans with survivors may also reduce patients' requests for unnecessary testing.
INTRODUCTION
There are nearly 14 million cancer survivors in the United States today.1 Earlier detection and treatment, combined with recent advances in treatment, contribute to survivors living longer with cancer. The number of cancer survivors is expected to increase to almost 18 million with the aging of the population in whom cancer incidence is highest.2 Combined with predicted shortages of primary care physicians (PCPs) and medical oncologists (MOs), these demographic changes present unique challenges for delivery of high-quality cancer survivorship care.3
The fragmented nature of the United States health care system presents additional challenges, highlighting the need for improved coordination and continuity of survivorship care. Patients receive care from many different providers (often located in geographically diverse areas) through several phases of illness over extended time periods.4 For patients more than 5 years after initial treatment, transitioning from an oncologist's care to follow-up care by a PCP requires care coordination that could potentially be improved by providing treatment summaries (TS) and survivorship care plans (SCP).5 Unfortunately, current physician reimbursement mechanisms do not support oncologists developing and delivering such plans. Studies have yet to demonstrate that plan use improves patient outcomes.6–7 Further, oncologists may hesitate to discharge patients to PCP care after monitoring patients closely over a long period and gaining patient trust.8 Yet with the growing number of new patients, oncologists simply cannot follow every patient throughout his or her remaining life span.
Insufficient communication between oncologists and PCPs can result in uncertainty among PCPs regarding appropriate post-treatment surveillance testing.9–11 The limited evidence base to support decisions regarding appropriate surveillance beyond 5 years after treatment presents an additional barrier to optimal care. Discrepant perceptions of MO and PCP survivorship care roles can also lead to missed or duplicated care.12 PCP and oncologist perceptions of their respective roles in survivorship care are more discrepant than with patient expectations of their roles.13 Patients may be satisfied with their care even if they perceive much greater participation by their PCP than the modest participation that may actually occur.14 Receipt of general noncancer preventive care is one benefit of concurrent follow-up by both MOs and PCPs compared with patients followed by oncologists alone.15 Other commonly cited noncancer-specific barriers to care, such as patient noncompliance, insurance restrictions impeding receipt of appropriate care, and language barriers, may also adversely affect patient receipt of appropriate survivorship care.11
Using nationwide survey data, we examined 10 self-reported barriers and concerns MOs and PCPs may encounter while caring for cancer survivors who are 5 or more years beyond completion of active treatment and the characteristics of physicians experiencing these barriers and concerns. To our knowledge, this is a novel analysis, entirely distinct from the general survey overview/methodology paper published recently.16 We hypothesized that MOs and PCPs would have different perceived barriers. The Mandelblatt et al11 adaptation of the behavioral model of access to care17–19 provides the conceptual framework for this analysis, as it incorporates communication between PCPs and oncologists in depicting realized access to cancer care.
METHODS
The Survey of Physician Attitudes Regarding the Care of Cancer Survivors (SPARCCS), cosponsored by the National Cancer Institute and the American Cancer Society, is a nationally representative mailed survey of United States PCPs and MOs conducted in 2009. The survey sought to improve understanding of the knowledge, attitudes, beliefs, and practices of these two groups of physicians regarding post-treatment care of breast and colon cancer survivors. Breast and colon cancer were chosen for study because of their high prevalence, long survivorship, and the availability of evidence-based guidelines for follow-up care.20–21 SPARCCS instruments are available on request.22 Study approval was obtained from the National Cancer Institute's institutional review board and the US Office of Management and Budget.
The complete survey methodology has been described elsewhere.16 Briefly, 2,525 PCPs (family medicine, internal medicine, obstetrics/gynecology specialists) and 2,750 MOs (oncology and/or hematology specialists) were sampled using a systematic stratified approach from the American Medical Association Physician Masterfile.23–24 Eligible physicians (n = 3,596) were required to practice in a nonfederal setting, be younger than 76 years, and spend 20% or more of their time caring for patients. Furthermore, oncologists had to provide care for breast or colon cancer patients within the previous year, and PCPs had to work in an office-based practice. Questionnaires from 1,072 PCPs and 1,130 MOs were considered evaluable. The absolute weighted response rate calculated using the American Association of Public Opinion Research RR3 method was 57.6% and the cooperation rate was 65.1%. Respondents did not differ significantly from nonrespondents. Survey weights were calculated using replicate jackknife methods.16
Dependent Measures
Of the 10 specific barriers/concerns examined in the survey, six were derived from existing literature.9,25 Four were designed by the survey team. Five were categorized as patient-related barriers or concerns (language issues, nonadherence to physician-recommended care, requesting more aggressive surveillance testing, inability to pay/lacking insurance,25 and insurance restrictions precluding test ordering25). The remaining five were categorized as physician-related barriers and concerns (ordering tests or treatments as malpractice protection, uncertainty regarding which physician [MO or PCP] is providing general preventive health care,9 concerns about duplicating care,9 missed care concerns,9 and inadequate knowledge or training to manage patient problems9). Respondents were asked how frequently they experienced each barrier (never, rarely, sometimes, often, always, or not applicable). To ensure an adequate number of respondents in each category to perform all analyses, “never” and “rarely” were collapsed into a single category, as were “often” and “always,” resulting in three levels. Two crude patient- and physician-related barrier summary variables were created by totaling all responses (never, 1; always, 5) across the patient- and physician-related barrier questions and converting the total to a dichotomous variable (< 15, low score; ≥ 15, high score).
Independent Measures
The primary independent measure was physician type—either MO or PCP. Physician characteristics, practice and patient characteristics, salary and reimbursement information, and self-reported knowledge and training were controlled for in the analyses. Physician characteristics included race/ethnicity, year of medical school graduation, United States or foreign training, and teaching status.
Practice and patient characteristics included primary practice site (eg, office or hospital) and location type (eg, full- or part-owner or employee of physician-owned practice), number of physicians in the primary practice location, metropolitan statistical area, census region, electronic medical records use, weekly patient volume, percent of patients categorized as “safety-net” (ie, Medicaid-insured or uninsured), and percent of patients 65 years and older. Salary and reimbursement included whether the physician was salaried and whether salaries were productivity based. Primary reimbursement was categorized as fee-for-service, capitation, or a mixture of both.
Several questions gauged physician confidence in cancer-related care knowledge regarding appropriate surveillance, long-term physical adverse effects of cancer and its treatment, and psychological outcomes. Because many individual knowledge questions were highly correlated, a summary variable was created (self-ascribed knowledge of follow-up care) and defined as tertiles of the distribution of summed knowledge question responses (high score, 17 to 18; medium, 14 to 16; low, 4 to 13). Training regarding late and long-term effects of cancer treatment was assessed, as were the training methods used in the previous 5 years (eg, continuing medical education or professional meetings).
Statistical Analyses
Analyses were performed using SAS (version 9.2; SAS Statistical Institute, Cary, NC). The surveyfreq procedure was used to apply replicate weights and conduct Rao-Scott χ2 tests. Weighted multinomial logistic regression models (surveylogistic with generalized logit link) were used to examine the relationship between each perceived barrier and the physician and practice characteristics. This technique permitted the use of polychotomous response categories (never/rarely, sometimes, often/always) for the dependent variables with never/rarely as the referent. All models included physician specialty (MO or PCP). Owing to the number of potential candidates for model inclusion, only variables with a Rao-Scott χ2 P value ≤ .15 were entered into each model, reducing the likelihood of overfitting and model noise. The Newton-Raphson method was specified for model fitting.
RESULTS
The 2,202 respondents (1,130 MOs and 1,072 PCPs) corresponded to a weighted total of 148,303 physicians (7,950 MOs and 140,353 PCPs). Physician type was significantly associated with all physician, practice, and patient characteristics; productivity-based salary; and all knowledge and training measures except training received from colleagues in the past 5 years (Table 1). Physician type (MO or PCP) was also significantly associated (P < .05) with all barriers to patient care in weighted-unadjusted analyses, except inability to pay/lacking insurance (Table 2). The average number of barriers and concerns physicians reported (often/always or sometimes) was 4.8 overall (standard deviation [SD], 2.2), 4.9 for MOs (SD, 2.2), and 4.7 for PCPs (SD, 2.3).
Descriptive Characteristics by Physician Type, Weighted Comparisons
Survivorship Care Barriers and Concerns by Physician Type, Weighted Comparisons
Patient-Related Barriers
For all respondents combined, the most common patient-related barrier was patient inability to pay/lacking insurance coverage (64.6%). Compared with PCPs in weighted-unadjusted analyses, MOs were significantly more likely to report that patients often/always request more aggressive surveillance testing than the physician would recommend (28.9% v 10.0%; P < .05), often/always/sometimes have language barriers that interfere with communication (25.7% v 20.3%), and sometimes have insurance restrictions that preclude obtaining necessary tests or treatments (43.2% v 37.3%). MOs were also significantly more likely to report that patients sometimes are noncompliant with recommended care (55.2% v 47.5%; P < .01; Table 2).
In weighted, adjusted, multinomial models, MOs were more likely than PCPs to report patients requesting more aggressive testing as a barrier (often/always: odds ratio [OR], 4.08; 95% CI, 2.73 to 6.10; sometimes: OR, 1.99; 95% CI, 1.53 to 2.58; Table 3). Compared with solo practitioners, physicians in small- to medium-sized practices (two to five and six to 15 physicians, respectively) were more likely to report patients sometimes requesting unnecessary testing. Physicians treating a higher percentage of safety-net patients (6% or more) were less likely to report patients sometimes requesting unnecessary testing as a barrier than were physicians treating few safety-net patients.
Patient-Related Barriers and Concerns to Cancer Survivorship Care: Weighted Adjusted Multinomial Logistic Regression Models
MOs were more likely to report patient language as a barrier to survivorship care compared with PCPs (often/always/sometimes: OR, 1.72; 95% CI, 1.22 to 2.42). Language was also a significant barrier to care among physicians (MOs and PCPs combined) in the northeast and west, those practicing at community health centers, in large practices (≥ 16 physicians) or in metropolitan statistical areas, and those reimbursed by capitation.
MOs were significantly more likely than PCPs to report insurance restrictions precluding the ordering of tests/treatments as a barrier (sometimes: OR, 1.42; 95% CI, 1.03 to 1.96). Insurance restrictions were also problematic for Hispanic physicians and those with productivity-based salaries. Conversely, respondents in medium-to-large practices (≥ six physicians), in the west, physicians trained in the United States, and those who were not paid by salary or whose salaries were not based on productivity were less likely to report insurance restrictions as a barrier to test ordering.
PCPs and MOs did not differ in reporting the remaining two patient-related barriers (patient inability to pay/lacking insurance or refusing recommended care) in adjusted analyses (results not shown). Although MOs had high patient-related barrier scores overall compared with PCPs in unadjusted analyses (P < .001), this association was no longer significant in adjusted analyses (P = .1132).
Physician-Related Barriers
For all respondents combined, the most common physician-related barrier was concern about missed care (55.7%). Compared with MOs in weighted-unadjusted analyses, PCPs were significantly more likely (P < .001) to often/always/sometimes report lacking adequate training (46.4% v 10.5%) and order tests/treatments as malpractice protection (51.1% v 40.2%). PCPs were also significantly more likely to often/always voice concerns about missed care (15.1% v 11.7%; P < .01). MOs were more likely (P < .01) to often/always/sometimes report concerns about duplicated care (56.0% v 47.7%) and general preventive health care responsibility (42.3% v 33.1%).
In weighted-adjusted multinomial models, inadequate knowledge or training to manage patient problems was significantly more likely to be reported as a barrier by PCPs compared with MOs (often/always/sometimes: OR, 3.06; 95% CI, 2.03 to 4.61; Table 4). Physician characteristics associated with a lower likelihood of reporting inadequate knowledge or training included black race, being in the midrange of the listed medical school graduation years (1980 to 1996), having high or moderate self-ascribed knowledge regarding follow-up care, and being trained in late- and long-term effects.
Physician-Related Barriers/Concerns to Cancer Survivorship Care: Weighted Adjusted Multinomial Logistic Regression Models
MOs were more likely than PCPs to report uncertainty regarding general preventive health care responsibility (often/always: OR, 1.97; 95% CI, 1.13 to 3.43; sometimes: OR, 2.16; 95% CI, 1.60 to 2.93; Table 4). Uncertainty was also more likely to be a concern for physicians (MOs and PCPs combined) in the west and those in practices that were transitioning to a full electronic medical record system. Physicians with high self-ascribed knowledge of cancer-related follow-up care and those physicians whose practice was 26% to 50% elderly patients were less likely to express uncertainty.
PCPs were more likely than MOs to report often or always ordering extra tests/treatments as malpractice protection (often/always: OR, 1.87; 95% CI, 1.20 to 2.93). Ordering extra tests or treatments was also more common among physicians who were reimbursed via capitation, were of Asian/Pacific Islander descent, saw more than 100 patients per week, and had only some late-effects training. Physician characteristics associated with a lower likelihood of ordering tests/treatments as malpractice protection included high self-ascribed knowledge of cancer-related follow-up care, detailed late-effects training, receipt of late-effects training from colleagues, being in the midrange of listed medical school graduation years (1980 to 1996), and having a salary not based on productivity.
PCPs and MOs did not differ in reporting the final two physician-related barriers (concerns regarding duplicated or missed care) in adjusted analyses (results not shown). Although PCPs had high physician-related barrier scores overall compared with MOs in unadjusted analyses (P = .0166), this association was no longer significant after adjustment (P = .9986).
DISCUSSION
Our article examined 10 self-reported barriers and concerns MOs and PCPs face regarding the care of breast and colon cancer survivors who are 5 years or more beyond completion of active treatment. To our knowledge, this is the first nationwide study of its kind. Our primary finding was that MOs and PCPs perceived different barriers and concerns as problematic, though the actual number of perceived barriers did not differ between MOs and PCPs (4.9 and 4.7, respectively). MOs were more likely to report concerns about language barriers when communicating with patients, insurance restrictions impeding ordering tests/treatments, patients requesting unnecessary testing, and general preventive-care responsibility. PCPs were more concerned with feeling inadequately trained and ordering extra tests/treatments as malpractice protection. Guidelines regarding cancer survivor follow-up care tend to be more explicit for the initial 5 years after treatment, but less so beyond 5 years.26–27 PCPs seem aware that there are additional services they should be providing for cancer survivors but may be unclear what those services should include.
Our findings regarding PCPs feeling inadequately trained in survivorship care and ordering excess tests/treatments suggest the need for improved physician education and training in survivorship care, particularly considering the growing population of long-term cancer survivors. Oncology and primary care professional societies should partner with cancer research organizations to develop improved guidelines and webinars for clinicians that address clinical knowledge gaps identified by provider studies such as SPARCCS and the unmet medical and psychosocial patient needs as reported by cancer survivor surveys.28–29 Guidelines should be expanded to include the period beyond 5 years after primary treatment and clearly differentiate among guideline-recommended care, care that may be considered but is not explicitly guideline-recommended, and care that is clearly not indicated.
MOs' concerns regarding requests from patients for unnecessary testing suggest a need for patient education. Patient navigators are trained to address education and two other concerns of MOs: language barriers when communicating with patients and insurance restrictions impeding necessary test/treatment ordering. Navigators can identify translators and link patients with support groups providing both educational and psychological structure. Navigators are also trained to assist patients in finding affordable care, often aided by social workers.30–31 The new standards for American College of Surgeons Commission on Cancer–accredited hospitals requiring that patient navigation be available either on site or by referral should help reduce oncologists' concerns.32
In addition, our findings regarding MOs' uncertainty about general preventive care responsibility suggest the need for increased coordination among clinicians in cancer survivorship care delivery. Survivorship care can be difficult to navigate. A team effort is needed to facilitate the patient-to-survivor transition.11 Improved care planning is needed beyond 5 years after primary treatment, despite the lack of survivorship care standards. For patients transitioning back to PCPs for continued care, TS and SCP are promising tools for improving continuity and coordination of care and for improving survivors' knowledge of optimal tests.33 Increased coordination among clinicians may also improve PCPs' confidence in their ability to manage patient problems. We found that PCPs who more frequently reported receiving a TS or SCP from the patient's oncology specialist were less likely to indicate having inadequate knowledge or training to manage patient problems (TS: often/always, 42.6%; sometimes, 41.9%; v rarely/never, 57.3%; SCP: often/always, 37.6%; sometimes, 44.5%; v rarely/never, 51.3%).
Potential study limitations include the focus on breast and colon cancer survivorship care barriers. Though the barriers included were fairly generic, it is possible that barriers encountered in providing follow-up care for children who are cancer survivors, for example, may differ from the barriers reported in this article. Similarly, as our analysis was comparative in nature, some barriers perceived solely by PCPs or solely by medical oncologists may have been overlooked in survey design. In addition, the self-reported nature of the data is a potential limitation. Over- or under-reporting is possible because of social desirability bias. For example, the number of physicians who reported ordering extra tests/treatments as malpractice protection may be an undercount of the actual number of respondents ordering more tests/treatments than necessary.
The importance of long-term follow-up care for cancer survivors should not be underestimated. Years ago, post-treatment survival for cancer patients was relatively short and surveillance was focused primarily on detecting recurrences and new primary cancers. With advances in diagnosis and treatment, patients are living longer and concerns about late effects of cancer and its treatment have come to the forefront, as have heightened concerns regarding psychosocial issues. The current study is important in highlighting potential challenges in transitioning survivors' care from MOs to PCPs and is unique in identifying MO- and PCP-specific concerns. High priority should be given to developing more efficient methods of communication among providers and between providers and patients,11 designing new models of survivorship care such as oncology medical homes,34–35 and establishing educational programs tailored to the specific concerns of PCPs regarding appropriate surveillance testing and late- and long-term effects of cancer and its treatment. Nationally representative physician surveys, such as SPARCCS, have been and will continue to remain crucial for systematic evaluation of the impact of such changes on cancer survivorship care.
AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST
The author(s) indicated no potential conflicts of interest.
AUTHOR CONTRIBUTIONS
Conception and design: All authors
Collection and assembly of data: Katherine S. Virgo, Catherine C. Lerro, Carrie N. Klabunde
Data analysis and interpretation: Katherine S. Virgo, Catherine C. Lerro, Carrie N. Klabunde
Manuscript writing: All authors
Final approval of manuscript: All authors
Footnotes
-
Supported by Grant No. HHSN261201000316C from the National Cancer Institute and American Cancer Society National Home Office Intramural Research Department.
-
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
