|Home | About | Journals | Submit | Contact Us | Français|
Little is known about how cancer physicians communicate with limited English proficient (LEP) patients. We studied physician-reported use and availability of interpreters.
A 2004 survey was fielded among physicians identified by a population-based sample of breast cancer patients. Three hundred and forty-eight physicians completed mailed surveys (response rate: 77 percent) regarding the structure and organization of care.
We used logistic regression to analyze use and availability of interpreters.
Most physicians reported treating LEP patients. Among physicians using interpreters within the last 12 months, 42 percent reported using trained medical interpreters, 21 percent telephone interpreter services, and 75 percent reported using untrained interpreters to communicate with LEP patients. Only one-third of physicians reported good availability of trained medical interpreters or telephone interpreter services when needed. Compared with HMO physicians, physicians in solo practice and single-specialty medical groups were less likely to report using trained medical interpreters or telephone interpreter services, and they were less likely to report good availability of these services.
There were important practice setting differences predicting use and availability of trained medical interpreters and telephone interpretation services. These findings may have troubling implications for effective physician–patient communication critically needed during cancer treatment.
There are growing numbers of limited English proficient (LEP) residents in the United States. According to the 2000 U.S. Census, 24 million or 9 percent of U.S. residents are LEP. The Census identifies LEP residents by first asking respondents if they speak a language other than English at home. Respondents who answer affirmatively are asked how well they speak English. Those who answer less than “very well” are designated as LEP (U.S. Census Bureau 2006). In Los Angeles County, 27 percent or 2.5 million residents are LEP (U.S. Census Bureau 2006). Among LEP Los Angeles County residents, 71 percent report speaking Spanish; however, there are populations of 1,000 or more representing more than 30 languages (U.S. Census Bureau 2000; Nichols et al. 2003;).
As LEP patients seek treatment in a predominantly English-speaking health care system, considerable gaps exist in knowledge about how physicians and health care organizations address linguistic barriers. The Institute of Medicine and other supporters of patient-centered care have advocated that patients should have the opportunity to participate in decision making about treatments and to communicate their symptoms and experiences to their clinicians regardless of English language proficiency (Davis, Schoen, and Schoenbaum 2000; Institute of Medicine Committee on Quality of Health Care in America 2001;). Studying how physicians caring for cancer patients address language barriers could be particularly useful because these patients typically require explicit education to partake of patient-centered care and optimally participate in treatment decisions (Kahn et al. 2002).
For example, women newly diagnosed with breast cancer frequently face a series of complex decisions regarding treatments in the face of symptoms and impaired quality of life. To make informed decisions, women typically need information and opportunities to meaningfully consider risks, benefits, and quality of life trade-offs. After treatment is initiated, women often need to engage with clinicians about their tolerance for treatments and side effects. Yet a number of barriers, including language concordance between patient and provider, constrain physician–patient communication (Sepucha et al. 2002; Epstein and Street 2007;). In this paper, we focus on language barriers between English-speaking physicians and LEP breast cancer patients.
LEP is reported to have a negative impact on access to care (Timmins 2002; Flores 2006; Ponce et al. 2006; Ponce, Hays, and Cunningham 2006; Yu et al. 2006;) and quality of care (Morales et al. 1999; Smedley, Stith, and Nelson 2003; Ayanian et al. 2005; Wilson et al. 2005; Chung et al. 2006; Flores and Ngui 2006; Ngui and Flores 2006; Divi et al. 2007; Karliner et al. 2007;). Studies have suggested that interpreters may improve quality of care for LEP patients (Kuo and Fagan 1999; Flores 2005; Green et al. 2005; Karliner et al. 2007; Kuo et al. 2007;). However, few studies characterize interpreter type, potentially blurring distinctions between interpreters (e.g., trained medical interpreter; telephone language interpretation services; bilingual staff not specifically trained in medical interpretation; and patients' friends or family members). Studies distinguishing interpreter by type have found that trained medical interpreters provide effective and efficient interpretation (Sarver and Baker 2000). Still, physicians using interpreters face certain challenges (Fagan et al. 2003; Karliner, Perez-Stable, and Gildengorin 2004; Gadon, Balch, and Jacobs 2007; Rosenberg, Leanza, and Seller 2007;), including longer visits (Tocher and Larson 1999; Kravitz et al. 2000; Fagan et al. 2003;) and inconsistent interpreter effects on care and outcomes (Sarver and Baker 2000; Green et al. 2005; Gadon, Balch, and Jacobs 2007;). Previous work with primary care physicians demonstrated that caring for LEP patients was more time consuming (Tocher and Larson 1999; Kravitz et al. 2000; Fagan et al. 2003;), unless the physician used a trained hospital interpreter (Tocher and Larson 1999).
This study reports results from a 2004 survey of physicians identified as providers by a population-based cohort of women newly diagnosed with breast cancer. We queried physicians about the proportion of LEP patients in their main practice (e.g., the proportion who do not speak English well enough to give an adequate history) and types of interpreters used to communicate with LEP patients within the last 12 months. This paper contributes to understanding how physicians communicate with patients in the presence of language barriers, and the availability of different types of interpreters when needed.
This cross-sectional study queried physicians identified as fulfilling key roles by a population-based sample of breast cancer patients in the Los Angeles Women's Health Study (LAW) (Chen et al. 2008; Yoon et al. 2008a,b;) regarding the structure and organization of breast cancer care (Tisnado et al. 2009). Physicians were considered as fulfilling key roles if they were named by a patient as specifically performing a cancer-related function such as being the doctor in charge of decision making for delivery of chemotherapy, radiation therapy, or surgery (Kahn et al. 2007). The sampling strategy is explained in greater detail in Appendix SA2. The research team obtained 348 mailed surveys from physicians associated with 298 unique office addresses (final response rate: 77 percent, 63 percent for medical oncologists, 88 percent for radiation oncologists, 75 percent for surgeons). After declining to respond to questions about LEP patients and interpreters, one physician was excluded. The sample for this analysis includes 111 medical oncologists, 65 radiation oncologists, and 171 surgeons (N=347). This study was approved by the RAND and UCLA Institutional Review Boards.
From the survey of cancer physicians in Los Angeles County, we selected three domains for analysis: proportion of LEP patients, use of interpreters, and availability of interpreters when needed.
Physicians were asked to identify the proportion of patients in their main practice who do not speak English well enough to give an adequate history, and to identify the language spoken by the majority of their LEP patients.
Physicians were queried about interpreters they used in communicating with LEP patients within the last 12 months. Survey respondents were allowed to indicate use of more than one type of interpreter. Response options included communicating without an interpreter because the physician was language concordant with the LEP patient, or using trained medical interpreters, telephone language interpretation services, bilingual office staff not specifically trained in medical interpretation, and/or patients' friends or family members as interpreters. The definitions of interpreter types represented by our response options are similar to those described by Hsieh (2006), Anderson et al. (2003), Downing and Roat (2002), and Tanjasiri (2001). However, physicians reporting use of trained medical interpreters or telephonic interpreters may not have had knowledge of the extent of training in medical terminology undertaken by either in-person or telephonic interpreters. Because trained medical interpreters can be in-person or available by phone, the trained medical interpreter category and the telephonic interpreters will not always be mutually exclusive. Cognitive interviews with these physicians indicated they understood the term “trained medical interpreter” as a way to categorize in-person interpreters, distinct from those accessible by phone.
Physicians were then asked how often each interpreter type (e.g., trained medical interpreter, telephone language interpretation service, bilingual staff not specifically trained in medical interpretation, patients' friends or family members) was available when needed. Response options included the following: always, often, sometimes, rarely, or never available when needed. Good availability was specified when a physician indicated an interpreter type was always or often available when needed.
Independent variables included self-reported physician demographics: age, gender, race/ethnicity and specialty (medical oncology, radiation oncology, and surgery), and practice characteristics: setting and size. The physician practice setting measure was derived from survey items about practice ownership and type. Practice settings were categorized as staff/group model HMO (reference group), county government or medical school/university, solo practice, single-specialty or multi-specialty medical group. Because within Los Angeles County government hospitals are also teaching hospitals, and because of small sample size, we treated county government and medical school/university as one practice setting.
Physicians reported the number of full-time physicians working in their main practice; large practice size was defined as any practice with 50 or more full-time physicians. The proportion of Medicaid or uninsured patients was not included in the multivariate analyses because this measure was highly correlated with the proportion of LEP patients (r=0.44, p<.000).
Univariate, bivariate, and multivariate analyses were performed for each dependent variable. The general linear model was used for analysis of proportion of LEP patients (McCullagh and Nelder 1989), and logistic regression was used for analyses of interpreter use, and availability of interpreters when needed (Hosmer and Lemeshow 1989). We tested for interactions between practice setting and large practice size, but no statistically significant differences were found.
Stata Version 10.0 (StataCorp, College Station, TX) was used to perform all analyses, weighting for nonresponse to avoid bias (Little and Rubin 2002). Survey nonresponse weights were calculated as the inverse of the probability of response based on a logistic regression model including physician gender, specialty type, study patient volume, and sharing an office with another surveyed physician. Medical oncologists and surgeons were found to be less likely to respond compared with radiation oncologists (p<.001 for medical oncologists, p<.05 for surgeons); and physicians who shared offices with another surveyed physician were more likely to respond compared with those who did not (p<.05). We controlled for clustering of physicians within office addresses (Wooldridge 2006). Using Hosmer–Lemeshow goodness of fit tests, all the logistic regression models appear to be a good fit to the data (p values range: .25–.91).
Univariate and bivariate results are shown in Table 1. Among the 347 specialists (e.g., medical oncologists, radiation oncologists, and surgeons), 99 percent reported caring for at least one LEP patient at their main practice within the 12 months before our survey. On average, physicians reported that almost one-fifth (17 percent) of their patients are LEP. However, the distribution was skewed; half of the physicians reported having fewer than 10 percent, while 10 percent of the physicians reported 35 percent or more LEP patients in their practices. Physicians most often reported that the majority of their LEP patients spoke Spanish (88 percent).
More than one-third (39 percent) of physicians reported communicating themselves in a non-English language with LEP patients. It should be noted that even physicians who reported communicating directly with LEP patients still reported at least one instance of using an interpreter to communicate with LEP patients within the last 12 months. Of physicians communicating with LEP patients without an interpreter, 46 percent reported speaking Spanish; others reported speaking Chinese (Mandarin or Cantonese), Japanese, Armenian, Farsi, Russian, and Korean. Although many physicians communicating without an interpreter reported proficiency in a language other than English, 30 percent did not.
Among those physicians who reported using interpreters to communicate with LEP patients within the 12 months before our survey, 41 percent reported using a trained medical interpreter; 21 percent reported using telephone language interpretation services; 76 percent reported using bilingual staff not specifically trained in medical interpretation; and 86 percent reported using patients' friends or family members for assistance in communicating with LEP patients. About one-third (32 percent) of physician respondents reported good availability of trained medical interpreters (e.g., they were always or often available when needed); and 33 percent reported good availability of telephone language interpretation services when needed.
After adjusting for physician race/ethnicity, specialty, and practice type, Hispanic and Asian physicians reported a higher proportion of LEP patients on average compared with non-Hispanic white physicians (coefficient for Hispanic: 0.81, 95 percent CI: 0.37, 1.25; coefficient for Asian: 0.36, 95 percent CI: 0.03, 0.69). Physicians in county government or medical school/university settings reported a higher proportion of LEP patients compared with physicians in solo practice after controlling for physician and practice setting characteristics (coefficient: 0.70, 95 percent CI: 0.14, 1.26). LEP patients appeared to cluster among Hispanic and/or Asian physicians, and physicians practicing in sites where free or low-cost care was offered.
Table 2 presents multivariate logistic regression results for use of interpreters within the last 12 months. We find that male physicians were less likely to report using telephone language interpretation services compared with female physicians (OR: 0.52, 95 percent CI: 0.27, 0.97). Hispanic physicians were more likely to report communicating with LEP patients without an interpreter (OR: 14.77, 95 percent CI: 3.13, 69.65), and less likely to use bilingual office staff not specifically trained in medical interpretation or patients' friends or family members compared with non-Hispanic white physicians (OR: 0.23, 95 percent CI: 0.08, 0.67; OR: 0.08, 95 percent CI: 0.02, 0.24, respectively). Radiation oncologists were more likely to report using telephone language interpretation services compared with medical oncologists (OR: 4.45, 95 percent CI: 2.01, 9.90).
Compared with physicians working in HMOs, physicians in solo practice and single-specialty medical groups were less likely to report using trained medical interpreters (for physicians in solo practice OR: 0.24, 95 percent CI: 0.09, 0.62; for physicians in single-specialty medical groups OR: 0.27, 95 percent CI: 0.12, 0.61, respectively). There were similar differences for telephone language interpretation services. Physicians in county government or medical school/university, solo practice, or single-specialty medical groups were less likely to report using telephone language interpretation services (OR: 0.31, 95 percent CI: 0.12, 0.78; OR: 0.30, 95 percent CI: 0.12, 0.71; OR: 0.21, 95 percent CI: 0.09, 0.48, respectively), compared with physicians working in HMOs.
Table 3 shows multivariate logistic regression results for good availability (i.e., physicians reported that the interpreter type was always or often available when needed). Radiation oncologists were more likely to report good availability of trained medical interpreters, telephone language interpretation services, or either, compared with medical oncologists (OR for trained medical interpreters: 3.80, 95 percent CI: 1.96, 7.35; OR for telephone: 6.46, 95 percent CI: 2.96, 14.10; OR for either: 4.57, 95 percent CI: 2.25, 9.32).
Compared with physicians in HMOs, physicians in county government or medical school/university settings, solo practice settings, or single-specialty medical groups, were all significantly less likely to report good availability of either trained medical interpreters or telephone language interpretation services. We find larger odds ratios (OR) for trained medical interpreters than for telephone interpretation services for physicians in county government or medical school/university settings (OR for trained medical interpreters: 0.36, 95 percent CI: 0.13, 0.99; OR for telephone language interpretation: 0.18, 95 percent CI: 0.07, 0.46), or in single-specialty medical group settings (OR for trained medical interpreters: 0.23, 95 percent CI: 0.09, 0.61; OR for telephone language interpretation: 0.18, 95 percent CI: 0.07, 0.46). For physicians in solo practice, we find similar ORs for trained medical interpreters (OR: 0.24, 95 percent CI: 0.09, 0.65) and telephone interpretation services (OR: 0.25, 95 percent CI: 0.10, 0.64).
There is a growing recognition of the importance of improving patient–physician communication and patient outcomes in cancer care (Epstein and Street 2007). Yet little is known about how doctors circumvent language barriers when communicating with LEP patients. We find that, although 99 percent of physicians in our sample reported having cared for LEP patients in the 12 months before our survey, less than half reported good availability of trained medical interpreters or telephone language interpretation services when needed.
Our analysis finds significant structural differences (e.g., staff/group model HMO practice setting and large practice size) in predicting use of trained medical interpreters; use of telephone interpreters; and availability of these interpreter types. Physicians and office managers from small group practices have reported that high perceived cost was an important barrier to arranging for language access services, and many reported little experience using and paying for such services (Gadon, Balch, and Jacobs 2007). Many physicians (62 percent) in our study are associated with small and solo practices and may face similar issues. In contrast, larger practice settings may have greater resources to offer trained medical interpreters or telephone services; they may also offer training to physicians and administrators on how to access and effectively use interpreters. Although the proportion of LEP patients was similar across practice settings, in larger practice settings, LEP patients may represent a larger absolute number of patients. Thus, larger practice settings may enjoy economies of scale sufficient to staff trained medical interpreters to assist physicians in their communications with LEP patients, or to contract with trained medical interpreter services at a lower per visit rate, as compared with smaller practices. Such organizations have the administrative capacity in place to identify and establish contracts with professional in-person and/or telephonic interpreter services, a level of administrative capacity that is likely lacking in many solo physician and small group offices. In addition, the staff/group model HMOs and large physician organizations are more likely to have the information technology capabilities to document and assess the prevalence of LEP patients in their enrolled populations.
Another possible explanation for the practice setting differences is greater levels of regulatory oversight. In Los Angeles, staff/group model HMOs and many of the network model managed care plans with whom larger physician organizations contract (Baumgarten 2005) may participate in Medicaid managed care and/or Medicare Advantage. These larger organizations may be more likely to be scrutinized for compliance under California Department of Managed Health Care regulations regarding language access (California Department of Managed Health Care, http://wpso.dmhc.ca.gov/regulations/docs/2009ccr.txt, accessed June 1, 2009; State of California, 1999) than physicians in solo or small practices. Thus, these larger practice settings may have greater incentives to maintain good availability of trained medical interpreters and/or telephone language interpretation services available when needed. However, we lack data to analyze these potential pathways.
By identifying practice setting differences in the availability and use of trained interpreter and telephone language interpretation services, we believe we have identified mutable structural differences that may present opportunities for interventions to provide LEP patients with greater access to interpreters. Because many practices lack a sufficient number of LEP to achieve economies of scale, Medicare reimbursement for interpreters could provide additional resources needed to encourage providers to hire professional interpreters when needed (Ponce et al. 2006).
We observe high rates of use, with no differences by practice setting, in the use of bilingual staff or patients' friends or family members. Other studies have found similar results regarding the prevalent use of patients' friends or family members as interpreters (Kuo and Fagan 1999; Gadon, Balch, and Jacobs 2007; Kuo et al. 2007;). Such frequent reliance on patients' friends or family members as interpreters may be a result of patient or physician preferences, or limited availability of trained medical interpreters or telephone language interpretation services.
The prevalent use of bilingual staff not specifically trained in medical interpretation and patients' friends and family members as interpreters is noteworthy since prior studies comparing the quality of interpreting by trained medical interpreters to that of patients' family members found that, although both groups made errors, errors of clinical consequence occurred more frequently with untrained staff and patients' friends or family members (Flores et al. 2003; Laws et al. 2004; Flores 2005; Karliner et al. 2007;). Additionally, untrained interpreters have a tendency to take on inappropriate roles as advocates for certain treatment options, rather than impartially conveying information (Preloran et al. 2005). Ideally, interpreters should facilitate communication without introducing their personal values; interpreters substituting their values or judgment for patients' is not consistent with articulated goals for patient-centered care. An additional clinical reality is that some LEP patients indicate a preference for having friends or family members interpret (Kuo and Fagan 1999). We concur with earlier work that advocated that patients' friends or family members should participate in decision making for patients who express such preferences (Matsumura et al. 2002). However, involvement by family or friends should not be considered a replacement for trained medical interpreters or telephone language interpretation services in facilitating important communication tasks including information exchange (e.g., patient history and symptoms, risks and benefits of different treatment options), discussing areas of uncertainty, making decisions, enabling patient self-management, responding to patient emotions, and fostering a healing relationship (Epstein and Street 2007).
A limitation of this study is that our analyses rely on physician self-report data. Because we asked physicians to recall interpreter use during the preceding 12 months, recall bias may have been an issue, though we have no reason to suspect systematic over or underreporting across practice types. Our study could not characterize the quality of the interpreter skills for physicians who reported communicating with their LEP patients without an interpreter. Additionally, this study has no measure of the quality of interpreters available to respondents, though evidence suggests there may be differences in effectiveness among different interpreter types (Flores 2005; Flores et al. 2003; Laws et al. 2004; Karliner et al. 2007;). Thus, we are unable to conclusively determine the quality or proficiency of trained medical interpreters or other interpreters.
It should be noted that our survey response categories were not worded to ask specifically about “professional, in-person” interpreters: the survey response category “trained medical interpreter” was used to represent this concept. A limitation of this study is the absence of standard national or state certification associated with health care interpretation. As a corollary, levels of training among professional interpreters, both in-person and telephonic, vary widely, and physician respondents may have been unaware of actual levels of training of professional in-person or telephonic interpreter services they used. We developed categories describing interpreter services used by physicians treating women with breast cancer in Los Angeles County following review of the literature and policy reports on health care interpreting (Kuo and Fagan 1999; Kravitz et al. 2000; Sarver and Baker 2000; Tanjasiri 2001; Downing and Roat 2002; Fagan et al. 2003;) after discussions with staff at community-based organizations in Los Angeles County involved with the training and provision of medical interpreter services to health care organizations throughout Los Angeles County. These discussions revealed concerns that interpreters in health care settings require facility with medical terminology as well as with conventional language. Because our main goal was to understand how physicians use interpreter services in the clinical setting, we chose a taxonomy that highlighted the term medical. With our survey, physicians reported on the availability of “trained medical interpreters; telephonic language interpretation services; bilingual staff not specifically trained in medical interpreting; and patients' friends or family members.” While we recognize that not all providers will be able to describe the training of their interpreters, physicians serve as excellent informants for reporting use and availability of interpreters who translate both medical terms and conventional language. A limitation is that these categories did not specifically ask about a “professional-in-person” interpreter; instead trained medical interpreter (which could have included some in-person and some telephonic interpreters) was used. We are, however, able to distinguish provider and practice setting characteristics that predict use of these interpreter types as compared with clearly ad hoc interpreters (e.g., bilingual office staff or patients' friends or family members), for whom it is reasonable to assume no specific training in medical interpretation.
Although this study inquired about availability of interpreters when needed, the study does not further characterize the elements necessary to access professional interpreters (i.e., knowledge of interpreter services, cost of interpreter services, challenges in hiring, or scheduling interpreter services). Further study is needed to identify additional factors, particularly cost, that may influence physician use of trained medical or telephone language interpreter services (Gadon, Balch, and Jacobs 2007).
This study does not specifically examine the impact of interpreters on LEP patients' satisfaction or outcomes, because we relied on provider self-report data. However, this report of providers' responses to communication barriers faced by LEP patients gives important new data on variations in use of interpreters of different types across practice settings, a topic not previously explored in the literature on physician report of strategies for communicating with LEP patients. This work will facilitate future interventions designed to improve communications with LEP patients by providing a foundation for understanding mutable structural aspects of care.
Physicians caring for LEP breast cancer patients in Los Angeles County report a variety of approaches in communicating with their patients. Although some report using trained medical interpreters and telephone interpretation language services, the overwhelming majority report using bilingual staff not specifically trained in medical interpretation and patients' friends or family members. Physicians appear to have limited access to professional interpretation services. The significant practice setting variations in the use and availability of trained medical interpreters and telephone language interpretation services imply that structural measures could facilitate use of these services by physicians caring for LEP patients.
A number of studies have shown that LEP patients have less access to care and less satisfaction with care. Our survey of providers may explain some of the factors leading to these disparities in care. Despite the complexity of treatment involved in breast cancer care, many providers reported using untrained interpreters to facilitate communication across language barriers even for complex decisions involving weighing risks, benefits, and quality of life trade-offs that require a full understanding of what treatments entail and what side effects may occur. We know that English-proficient women often struggle to effectively communicate with physicians during cancer treatment (Sepucha et al. 2002). It is thus not difficult to imagine LEP women facing even greater challenges in fully understanding their treatment options and in communicating their hopes, fears, and experiences to their providers. A recently published research agenda for language barriers in health care (Jacobs et al. 2006) noted more research is needed on how physicians can use interpreters to improve communication. This analysis makes clear that physicians have an important message about lack of availability of interpreters when needed for cancer patients. In response, physicians, patients, and policy makers can consider strategies to ensure that physicians in all settings effectively address the needs of LEP patients to achieve better communication outcomes as a prerequisite to patient-centered cancer care. These include recommendations for reimbursement for interpreter services; establishment of uniform standards for health care interpreting; clarification of existing regulations regarding the use of interpreter services in improving language access; and training for physicians on the use of professional interpreters while caring for LEP patients. In addition, the establishment of quality indicators with respect to patient–physician communication would be a major step toward achieving better communication outcomes necessary to patient-centered cancer care.
Joint Acknowledgment/Disclosure Statement: Dr. Rose received support from the UCLA Cancer Education and Career Development Program, National Cancer Institute grant R25 CA087949. Dr. Tisnado received support from a U.S. Department of Defense Breast Cancer Research Program Postdoctoral Award DAMD17-03-1-0328 and The University of California Los Angeles Resource Center for Minority Aging Research/Center for Health Improvement of Minority Elderly (RCMAR/CHIME) NIH/NIA grant P30-AG021684. This research was funded by the California Breast Cancer Research program grant 7PB-0126S and the National Cancer Institute and the Agency for Health Care Research and Quality grant 1-R01-CA81338-01A1. We thank the two anonymous referees for suggestions that significantly strengthened the analysis.
Disclaimer: The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs or the United States Government.
Additional supporting information may be found in the online version of this article:
Appendix SA1: Author Matrix.
Appendix SA2. Description of Sample for the Los Angeles Women's Health Study Provider Survey.
Please note: Wiley-Blackwell is not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.