|Home | About | Journals | Submit | Contact Us | Français|
Agreement between patients and physicians is an indicator of successful communication. Concordance in domains of communication among patients with heart disease and communication barriers has not been studied.
English, Spanish, or Cantonese-speaking patients seen at a public hospital cardiology clinic were assessed with pre-visit questionnaires. Surveys of patients and their physicians immediately after the visit asked each about: (1) cardiac functional status, (2) barriers to self-management, (3) cardiac diagnoses, and (4) treatment. We assessed patient-physician concordance in these domains.
179 patients and 56 physicians completed the study. Patients had low educational attainment, limited literacy and limited English proficiency. Physicians underestimated patients’ cardiac functioning status (NYHA Classes 2–4), by 1 class or more in 50% of visits. Physicians were frequently unaware of medication (38/57, 67%) and psychosocial (61/88, 69%) barriers. Patients were unable to describe even 1 matching diagnosis (72/170, 42% concordant) among 5 categories. Physicians’ reported medication changes in 106/179 (59%) but patients failed to report these changes in 55% (58/106). Multivariate logistic regression analyses showed no significant association between patient characteristics and concordance.
Patients and physicians often fail to communicate effectively and determinants of concordance in CVD care require further investigation.
Developing strategies to improve communication within the medical encounter are critical to improving ambulatory chronic disease management.
Over 80 million Americans have cardiac disease, making it a leading cause of morbidity and mortality in the U.S.. Moreover, addressing cardiovascular disease (CVD) disparities is a major public health focus . Managing chronic cardiac conditions requires extensive use of ambulatory health care services. In turn, patients with chronic cardiac conditions must perform complex self-management in order to avoid complications . While enabling and supporting this self-management requires clear communication between patient and physician, little is known about patient- physician communication in ambulatory CVD care. The Institute of Medicine has concluded that high quality of care is the product of good interpersonal processes of care, including communication, combined with good technical processes of care [5, 6]
Similar to other chronic disease conditions, information exchange is necessary for effective ambulatory cardiac disease care. Schillinger developed a conceptual framework for communication needed for effective chronic disease management, and how health literacy affects distinct domains of communication . We adapted this framework to describe communication processes in ambulatory health visits (see Figure 1, adapted from .). In a typical ambulatory encounter, the physician needs to elicit the patient’s current symptom severity as well as barriers to treatment adherence, including psychosocial factors. For a heart failure patient, for example, this may include the information that the patient has become more short of breath and/or has recently run out of medication at home. This elicitation-type communication provides information needed for decision-making. Knowing the patient’s clinical status and behaviors, the physician can make informed clinical recommendations to the patient about the treatment plan and engage in shared decision-making. The next phase of the ambulatory encounter requires adequate explanatory-type communication. Here, the physician must clearly convey diagnostic information to the patient, and her recommendations to the patient so as arrive at a shared decision regarding the treatment plan. The Institute of Medicine has defined patient-centered care as providing care that is respectful of and responsive to individual patient preferences, needs, and values and ensuring that patient values guide all clinical decisions . Clearly, achieving patient-centered care requires active elucidative as well as explanatory communication . In the case of the heart failure patient, the physician might explain heart failure and its cardinal symptoms, provide a medication prescription or alter a dose, and advise close follow-up if symptoms worsen. Adequate explanatory communication is needed so that the patient can self-manage his condition, adhere to treatment, and thereby avoid a poor health outcome. For a patient who does not recognize worsening symptoms, or does not take medications correctly due to lack of understanding, his heart failure could worsen and require acute care.
We applied this model to investigate communication within the ambulatory cardiology encounter in an ethnically and linguistically diverse population. We report the extent of concordance, or agreement, between patient and physician for (a) elicitation of patient’s health state, medication barriers and psychosocial barriers; and (b) explanation of treatment plan, including medications, diagnoses, and clinical options. In addition, we explored whether patient characteristics were associated with concordance in each of these domains while prior studies in cardiac care have measured concordance in one domain [10, 11], To our knowledge none has simultaneously measured the 4 domains of health state, medication and psychosocial barriers, medications prescribed, and diagnoses, and few focus on populations known to experience disparities in outpatient cardiac care .
We performed this study in the cardiology clinic at Alameda County Medical Center (ACMC), a safety net setting which predominantly serves uninsured and publicly insured ethnic minorities. The clinic met on Mondays and Thursdays, and was staffed by two attending cardiologists. Resident physicians training in internal medicine rotate through the clinic and see patients under the supervision of the cardiologists. All visits carried out by residents were presented to attending physicians in real time, and the attending also generated a chart note. Language interpretation for visits involving non-English speakers and language discordant physicians includes in-person, telephone, and real-time videoconferencing. This communication study was nested within a larger observational study of language interpretation methods.
We approached all English, Spanish, and Cantonese-speaking patients 18 years and older who were being seen at the cardiology clinic of ACMC in Oakland, CA between March 2004 and January 2005. Because we were interested in exploring the effects of limited English proficiency, we aimed to maximize inclusion of non-English speaking patients. Therefore, all Spanish and Cantonese speaking patients were approached for potential recruitment, and a similar number of English-speaking patients were recruited in parallel. All patients answered standardized questions assessing comprehension of the informed consent, and those who could not correctly answer the 4 post-consent questions were provided an additional round of explanations, and if the 4 post-consent questions were not correctly answered, these potential participants were excluded on the basis of presumed cognitive impairment or significant communication barrier. We invited all clinic physicians to participate in the study. The study was approved by the Committee for Human Research at the University of California, San Francisco and the Institutional Review Board at ACMC.
We conducted a brief, in-person, interviewer-administered baseline assessment for all participants at the time of enrollment, just prior to their visit. All measures were translated into Spanish and Cantonese from English versions and back-translation was performed to confirm accuracy. The baseline assessment included a demographic and health status survey which included age, gender, educational attainment, and race/ethnicity (White, African American, Latino, Chinese). We used the general health question of the SF-12 version 2 , “In general, would you say your health is” (excellent, very good, good, fair, poor)” to assess health status. Patients also reported whether they usually had an interpreter at cardiology visits. English and Spanish speakers underwent the short-form Test of Functional Health Literacy Assessment (s-TOFHLA).
Following the visit with the physician, research assistants administered a post-visit questionnaire to patients and treating physicians completed a written, self-administered post-visit survey with parallel items. The physician who saw the patient completed the survey. For visits in which the resident and the attending physician both examined the patient, the resident completed the survey, as the resident generally spent more time with the patient than the supervising attending physician. First, we measured how heart disease affected the patient’s functioning, as follows: “Which one of the following statements best describes how your heart problem affects you today?” The four response options corresponded to the New York Heart Association Class I (no cardiac symptoms even with activity); Class II (cardiac symptoms lead to some difficulties with usual activity); Class III (cardiac symptoms make patient unable to perform usual physical activity); or Class IV (cardiac symptoms present even at rest) . This classification scheme has been used to identify differing functional status among populations with cardiac conditions, even beyond those with heart failure [16, 17]. Second, we asked about specific barriers to medication adherence over the prior 4 weeks: running out of medication, inability to pay, concern about side effects and effectiveness, and lack of understanding of medication instructions. Third, we inquired about several psychosocial barriers to self-care over the prior 4 weeks, including problems with finances, family, housing, or pain . Fourth, we asked patients to tell us whether any medication changes were made during the visit. We simply asked patients to report whether and how many medications were started, stopped, increased, and/or decreased. We did not ask patients for names or doses of medications. Physicians were asked to write down the name and dose for each medication initiated or discontinued, any changes to the standing medication doses during the encounter. Fifth, patients were asked to report, in their own words, up to 3 heart problems or diagnoses related to the visit. Physicians were asked to report up to three cardiac diagnoses for each patient. In order to obtain the most complete data, for encounters with a missing physician post-visit survey, two study investigators (KBD, DS) also performed chart abstraction to obtain medication changes and diagnosis information.
In assessing for concordance, we distinguished between elicitation and explanatory communication domains. In this framework, patients’ reports of their own cardiac functioning and barriers to self-management (elicitation-type communication) were considered complete and accurate (i.e. the gold standard), and so we deemed physician responses as either concordant or not. In contrast, we considered physician reports (or, for those visits that lacked a physician report, we abstracted the information from the written clinical record) as accurate (i.e. the gold standard) for explanations of cardiac diagnoses and medication changes (explanatory-type communication), and examined whether patient reports matched with their physician’s reports.
Because patients’ descriptions of diagnoses rarely used medical terminology, we employed a coding scheme used in prior research  that listed medical concepts with their equivalent lay terms for application of systematic coding categories across visits. As an example, if a physician reported that the patient had congestive heart failure and the patient reported that their “heart doesn’t pump well”, we considered these to be concordant (a list of concordant and discordant diagnoses is available from the authors upon request). We pre-defined medication change concordance such that if patients reported the same count for each type of medication change as their physicians, they were considered concordant (i.e., if the physician reported stopping warfarin and the patient reported stopping 1 medication, they would be concordant on medication stops).
We defined several specific concordance measures to employ in exploratory multivariate analysis of correlates of concordance. For symptom concordance, we considered a difference of 2 or more NYHA Classes to be discordant, because this difference is clinically significant . For medication barriers, we selected the most common medication barrier-- concern about side effects --- as a representative item. Similarly, we chose one psychosocial barrier --problems with pain --- to test in the multivariate model. For medication changes, if the patient reported a different number of medications started, stopped, increased, and decreased than did the physician, we considered that type of medication change to be discordant. Finally, we considered diagnoses concordant if the patient matched on at least 1 of the physician diagnoses, even if they did not match on others.
We first report summary statistics showing the extent of concordance for current functioning, barriers to self-management (individual medication-use and psychosocial barriers), medication changes, and diagnoses. Not all physicians and patients answered all questions in the survey, leading to some missing values for concordance outcomes. Item non-response is a common problem in surveys, and was a particular concern for us with physician surveys. Analysis restricted to complete cases can introduce bias if the data are not missing at random, and also reduce precision . Using standard methodology, we imputed all cases in which we had patient baseline data . Therefore, we used multiple imputation, as implemented in SAS Proc MI . This procedure uses the distribution of existing co-variates to estimate a distribution of values for missing variables, in multiple iterations, to generate 20 completed datasets. Imputed values for binary and categorical variables were rounded and truncated to the nearest category . Final parameter estimates, confidence intervals, and significance tests were calculated using standard methods for combining results across the 20 imputed data sets [23, 24] as implemented in Proc MIANALYZE.
To explore whether any measured patient factors were associated with concordance in the pre-specificied indicators across communication domains (functioning; 1 selected medication barrier- concern about side effects; 1 selected psychosocial barrier-problems with pain; medication changes; diagnoses), we created separate multivariate logistic regression models, using the imputed datasets as above. We pre-specified the co-variates for all multivariate models as follows: age, language (English, Spanish, or Cantonese), race, use of language interpreter, educational attainment, and number of cardiac diagnoses as determined by the physician. For each model, we only included the participants with a positive response for that communication domain. For instance, the medication changes model included only the participants for whom medication changes were made at their visit. We chose this approach because we felt that concordance on the absence of an action or barrier would differ significantly from concordance on the presence of an action or barrier. We expect clinician’s lack of awareness of barriers to be more likely to interfere with appropriate treatment strategies than agreement about a lack of barriers.
Of 269 patients approached, we enrolled 179 patients (67%) treated by 56 physicians, including both resident and attending physicians. Eighty-five patients (31%) refused, and 5 (2%) could not demonstrate understanding of the informed consent. For 107 (60%) of the encounters of enrolled patients, physicians completed a post-visit survey; abstracted chart data regarding diagnoses and medication changes were obtained for 58 (80%) of the remaining encounters. For every case, there was a patient baseline survey. However, in 15 cases physician responses were completely imputed, and in 2 cases the patient post-visit responses were imputed.
The patient population was ethnically diverse, with overall low educational attainment and a high prevalence of limited health literacy (Table 1). Forty-two percent were English speakers, 43% Spanish speakers, and 15% Cantonese speakers. The population had a significant illness burden, with 124 (69%) reporting either fair or poor health status.
First, with respect to cardiac functional status, physicians significantly underestimated patients’ NYHA class (Figure 2). For example, among the 19 participants with cardiac symptoms even at rest (NYHA Class IV), physicians were discordant in 16 of 19 cases (82%); the magnitude of this difference was at least 2 NYHA Classes in 14 of 19 (74%) cases. Second, physicians were commonly unaware of patients’ medication barriers (Table 2). The most common medication barrier for patients was concern about side effects, but their physicians were concordant in identifying this as a patient barrier only 23% of the time. Third, patients’ psychosocial barriers almost always went unrecognized, with concordance ranging from only 7–18% of cases (Table 2).
Discordance was also present in explanatory-type communication domains. First, we found that patients’ ability to report diagnoses varied, ranging from 22% for hypertension to 63% for valvular heart disease (Table 3). In addition, patients reported the correct number of medication changes 52–67% of the time, depending on the type of change (Table 3). Overall, for 59% of patients, physicians reported medication changes. Of these, only 46% of patients reported complete concordance with physician reports.
In our exploratory multivariate analyses, we did not find consistent or significant associations between concordance in any communication domain and age, language, race/ethnicity, use of language interpreter, educational attainment, or number of cardiac diagnoses (Table 4). Among English and Spanish speakers, we did not find an association between health literacy level and concordance, when adjusted as above for age, language interpretation, race/ethnicity, and number of cardiac diagnoses (data not shown).
In a diverse, largely uninsured population with chronic cardiac disease cared for in a public hospital cardiology clinic, we found a marked lack of concordance across multiple communication domains, especially in physician elicitation of patient concerns. Concordance, or agreement between patient and physician about the patient’s diagnosis, patient’s health status and treatment plan and treatment/self-care barriers, is a crucial indicator and outcome of high-quality communication [10, 25, 26]. Poor concordance between patient and physician regarding the standing medication regimen has been reported in a number of studies conducted in a variety of ambulatory settings, but mainly in primary care [10, 27–32]. This suggests that poor communication is common and widespread – a problem that can be greatly exacerbated when there are frequent medication changes and cultural and language differences between patients and physicians . Recent studies suggest that limited health literacy and limited English proficiency jeopardize many domains of patient-physician communication, even in the presence of language interpretation or language-concordant physicians [34, 35]. Perhaps because communication difficulties were so widespread, our exploratory analyses did not identify patient characteristics that were independently associated with worse concordance.
Poor concordance has been linked to unfavorable outcomes, including decreased patient satisfaction [36–38], worse adherence to future appointments [36–40], and some health outcomes [41–44]. As an example, one recent study in anticoagulant care demonstrated that concordance for medication regimen was associated with fewer medication errors and better disease control.
Our study was set in an ambulatory sub-specialty clinic, in which communication has not been well characterized. Similarities to the primary care setting include care of chronic illness, the structure of the short office visit every few months, and the emphasis on medication use. However, the cardiology encounter does not have to address the breadth of issues that a primary care visit does; thus, one would expect a narrower focus of the communication in the cardiology setting than in primary care. We would expect the focus on a single body system to result in greater concordance.
We found significant impairment in elicitation-type communication. As expected in a setting that provides care to vulnerable populations, a significant proportion of patients reported psychosocial barriers, and physicians were often unaware of these factors. Similar findings have been demonstrated in other clinical settings, including those that care for insured patients [45–47]. Insofar as patients may not report medication and/or psychosocial barriers to their physicians because of social desirability, social distance and/or concerns related to stigma, our findings suggest that physicians need to ask and be aware of the social factors that affect their patients’ self-management. This is particularly critical when working with patients that have multiple barriers to self-management. Communication techniques that can help uncover such sensitive topics have been shown to be effective [48, 49], and may be useful in ambulatory cardiology practice. Similarly, communication training that enables patients to be more active participants in visits can improve communication and even health outcomes . Importantly, this study included resident physicians, and their lack of clinical experience may have contributed to poor elicitation-type communication. Elicitation-type communication is critical to patient-centeredness, which in turn is fundamental to quality medical care .
Similarly, explanatory communication was also sub-optimal. For instance, we find it alarming that only 15% of those with hypertension could report having it as a diagnosis, since those who do not recognize themselves as having a condition may be less likely to adhere to medications and lifestyle recommendations. Those patients with valvular heart disease were most likely to report it, which may reflect either the long duration or severity of the condition, or the relative ease with which valvular disease can be explained.
While we asked physicians and patients about medication changes immediately following the encounter, patients nevertheless, often could not correctly report medication changes. While poor recall of medication changes has been previously demonstrated , this is exacerbated by the ever-expanding number of medications used in ambulatory cardiology practice.
Our study has notable strengths, including its inclusion of multiple communication domains, the very diverse patient population, and its setting in an ambulatory cardiology practice. However, it has a number of limitations. First, while communication is particularly important in the context of limited health literacy and English proficiency, our results may not generalize to more affluent and educated populations cared for in non-teaching environments. However, a study among congestive heart failure patients found significant medication discordance even in an English-speaking, community-based, insured sample. Second, the modest sample size limited our ability to construct complex multivariate models with large numbers of potential predictors and adjust for within-provider clustering. Third, not all physicians completed the post-visit survey. We addressed this at the time of data collection by abstracting as much information as possible from the chart; we acknowledge that providers may have had verbal communication about medication use, barriers, and psychosocial barriers that they did not include in chart documentation. To address missing data in the analysis stage, we used multiple imputation techniques, a preferred technique to address non-response bias . Fourth, we do not have information about the proportion of physician-respondents who were trainees. Fifth, we did not ask patients to report the names and doses of their new medications, but instead the count. We chose this approach because prior studies have shown that physicians do not always tell the patient the name and dose of a new medication when it is prescribed [31, 32], but even with a simple count, concordance was no better than 50%.
Overall, patient-physician concordance in both elicitation and explanatory domains of communication was sub-optimal in this study of socially vulnerable adults with CVD. Even in a setting with consistent professional language interpretation, physicians were frequently unaware of patients’ functioning status and barriers, while patients frequently did not come away with a clear understanding of their diagnosis or treatment plan.
This pervasive medication regimen discordance underscores the importance of real-time ambulatory medication reconciliation, using visual as well as written display [52–55]. Moreover, regimen discordance has the potential to cause adverse drug events, and may partly explain the high prevalence of CVD medications as causes of adverse drug events leading to emergency department visits .
As in prior studies [17, 57], we found that physicians underestimated patient cardiac functional status, and this effect was most pronounced among those with the most functional limitation. This under-recognition is particularly concerning, because patients’ NYHA class typically guides cardiologists’ treatment . Moreover, physicians were frequently unaware of patients’ barriers to medication use. Some of these barriers, including concern for side effects and effectiveness, as well as limited understanding of regimen, might directly be addressed in the ambulatory visit were the physicians to have elicited them. Prior studies have shown that medication counseling can improve patient regimen understanding and adherence [10, 59, 60].
These findings underscore the need to improve communication in ambulatory specialty settings, particularly for those with limited health literacy and limited English proficiency. Standards and certification for non-English language fluency , and medical interpretation may be useful and certainly merit further study. Moreover, previous research with limited health literacy populations has demonstrated several effective communication techniques, which may be useful in improving concordance . For example, the “teach-back” method, in which providers ask patients to re-state key information, has been associated with better metabolic control among patients with diabetes , improvements in self management among asthma patients , and increased comprehension of informed consent information , regardless of health literacy. In addition, communication aids such as visual medication schedules have been shown to improve patients’ understanding, reduced medication errors, and, in the case of oral anticoagulation, improve outcomes [52–55]. Evidence-based communication strategies such as these should become part of routine clinical care for ambulatory cardiac patients, especially in safety-net settings.
Future research should examine both predictors of poor concordance in ambulatory CVD care, to determine which patients are at highest risk for poor communication. Moreover, possible adverse health outcomes associated with discordant communication should be studied, particularly in ambulatory CVD.
This study was supported by grant no. 20061003 from The California Endowment and by grant no. P30-AG15272 of the Resource Centers for Minority Aging Research program funded by the National Institute on Aging, National Institutes of Health. We thank Dana Nickleach and Ginny Gildengorin for data analysis support, Steven Gregorich for statistical advice, Gabriel Somma, Monica Lopez and Julissa Saavedra for data collection and management, Andrea Lopez and Cecilia Populus-Eudave for assistance in manuscript preparation and submission, and the staff and physicians at the Alameda County Medical Center Cardiology Clinic for their participation.
Dr Sarkar was supported by Agency for Healthcare Research and Quality K HS01759; Dr Schillinger was supported by NIH Clinical and Translational Science Award UL1RR02413 (CTSI grant). This work was also supported by the National Center for Research Resources (KL2 RR024130).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.