Physician Role Obligations and Medicine's Unwritten Social Contract

It has been argued that the basis of trust between patients and their physicians lies in the physician's dedication to "universalism," that is, the responsibility to treat all patients alike without regard to particular attributes or ascribed traits (Parsons, 1951). It is reasoned that if patient care is not universalistic, suspicion and caution will prevail over trust and confidence in the patient-physician relationship. Fear that physicians might act upon ageist, class, or racist stereotypes could undermine the fabric of the social contract upon which the therapeutic relationship rests. In light of the significance of potential violations of physician universalism, investigation of the association between patient attributes and aspects of care should be a research priority. However, this has not been the case. There have been relatively few methodologically sound studies designed specifically to investigate the role of sociologic factors in medical visits(Greene, Adelman, Charon, & Hoffman, 1986; Gerbert, 1984; Roter, Hall, & Katz, 1988 Roter & Hall, 1992).

The Nature and Consequence of Broad Normative Expectations, Bias, and Racial Stereotyping by Providers and Patients

There are three mechanisms by which one might hypothesize physician behavior would relate to patient characteristics (Roter & Hall, 1992). First, there may be an unintended association between the care process and patient attributes that is produced by mutual ignorance of social or cultural norms. The marked differences that often exist between physicians and their patients (for example, patients who are poor, uneducated, and belong to an ethnic or racial minority group) may lead to very basic communication difficulties. For instance, citing sociolinguistic theorists, Waitzkin (1985) has generalized to the medical context the finding that middle-class subjects tend to be verbally explicit, while working-class subjects tend to communicate more implicitly through nonverbal signals. If not attuned to these nonverbal signals, physicians could easily miss or misinterpret patient requests for information or reassurance.

A second explanation for an association between patients' sociodemographic characteristics and the medical care process is that physicians may be consciously and quite appropriately addressing the varying responses to illness demanded by socially patterned expectations for care. . These needs reflect the diverse attitudes, beliefs, and expectations of the groups to which the patients belong (Fox & Storms, 1981). For instance, in his classic study of ethnicity and pain, Zborowski (1952) found that patients' interpretation of pain and expectations regarding pain control varied widely across ethnic groups and that members of these groups communicated these expectations to their physicians. In these instances, effective tailoring of pain management maximized medical care.

Finally, it is possible that physicians, like others in our society, are negatively affected by stereotypes. Physicians have generally scored about the same as non-physicians in surveys reflecting attitudes toward the elderly or the poor (Marshal, 1981; Price, Desmond, Synder, & Kimmel, 1988). Further, the range in physicians' political and ideological beliefs indicates a broad spectrum of response to patient groups (Waitzkin, 1985). Physicians appear to share the same negative stereotypes about physically unattractive people as do others (I know this changes the meaning—but is it really fair to say that **everyone** has negative stereotypes about unattractive people?) in our society (Nordholm, 1980). Numerous studies indicate that patient race and ethnicity influence physicians' beliefs about and expectations of patients (Lewis, Croft-Jeffreys, & David, 1990, van Ryn & Burke, 2000, Porter & Beuf, 1994, Schulman et al., 1999). One recent study used survey data from patients and physicians during post-coronary angiogram encounters to examine the effect of patient race and socioeconomic status (SES) on physician perceptions of and attitudes towards patients(van Ryn & Burke, 2000). This study showed that even after adjustment for patient age, race, frailty/sickness, depression, mastery, social assertiveness, and physician characteristics, physicians tended to perceive African Americans and members of low and middle SES groups more negatively on a number of dimensions than they did whites and upper SES patients. For example, African-American patients were perceived as being less intelligent, more likely to engage in high-risk behavior, and less likely to adhere to medical advice. These ethnic minority patients received lower ratings of affiliation by physicians. Patients in the lowest SES group were also perceived as having more negative personality attributes (lack of self-control, irrationality), less abilities, more negative behavioral tendencies, and fewer role demands (van Ryn and Burke, 2000). Patient SES appeared to have an even broader effect on physician perceptions than patient race. These perceptions could not be completely explained by epidemiologic evidence about the patients' racial or SES group or from patients' reports of their actual behaviors and tendencies. Physicians' negative attitudes or the assumptions they make about a patient's personality, motivation, or level of understanding clearly have implications for the care they give.

Correlates of Communication

Individuals coming together in medical dialogue bring with them all of their personal characteristics, including their personalities, social attitudes and values, race, ethnicity, gender, sexual orientation, age, education, and physical and mental health. . This applies to the physician as well as to the patient, though research on physician characteristics is less common, owing to typically small physician samples in communication studies(Hall & Roter, 1992). Furthermore, the endpoints we might wish to measure, such as satisfaction or clinical outcomes, have many determinants. When interpreting nonexperimental comparisons such as that comprising most of the literature on physician-patient communication, it is important not to make assumptions about the causal relations among variables. Even when potentially confounding variables (such as sociodemographic variables or health status) are statistically controlled, strong inferences of causality are often not justified. Causation may lie in variables unrelated to those under study, and even when one has measured the right variables, complex paths of causation can exist. In the medical interaction for example, mutual (reciprocal) influence is possible. Although a given behavior may be produced by a quality of the person engaging in it, it may also be caused by how that person responds to the other person, or how that person is treated by the other. . Therefore, while we tend to think of physicians as the active agents and patients as responding to them, influence does not always flow from the physician to the patient (Roter & Hall,1992) and that physicians are influenced by patients, too. For example, one recent study shows that the effect of patients' race or ethnicity on physicians' beliefs about patients can be moderated by patient behavior. Krupat and colleagues (1999) conducted a study in which physicians observed randomly assigned videotapes of women seeking care for breast cancer. The videotapes varied patients' sociodemographic characteristics, general health status, and assertiveness. The study showed that assertive behavior among black and low SES patients, but not in white or upper SES patients, resulted in a greater likelihood that physicians would order full tumor staging.

In the sections that follow we present research relating physician-patient communication to some of the antecedent and outcome variables that are relevant for understanding and eliminating racial and ethnic disparities in healthcare.

The Role of Physician Sociodemographic Characteristics on the Medical Dialogue

Physician race and ethnicity . Few studies have explored the impact of physician race and ethnicity on medical communication. Most of these studies have used patient ratings of the quality of the patient-physician relationship and the physician's communication style, rather than actual measures of communication, such as audiotape, videotape or direct observation. A study of 1816 adults and 64 primary care physicians in a large managed care organization in the Washington, D.C. metropolitan area examined how race and gender of physicians and patients were associated with patient ratings of physicians' participatory decision-making (PDM) style. The physician sample was 56% white, 25% African American, 15% Asian American, and 3% Hispanic. This study showed that there were no differences between ethnic minority and white physicians with respect to patient ratings of PDM style, even when adjustments were made for patients' age, gender, education, health status, and length of the patient-physician relationship (Cooper-Patrick et al., 1999). This finding is in contrast to the Medical Outcomes Study, in which nonwhite physicians were rated as less participatory than white physicians. In this study, the ethnic mix of the nonwhite physician group was not reported. It is therefore unclear which cultural or ethnic factors contributed to the lower participatory ratings by patients and whether the lower participatory ratings can be attributed to one specific ethnic or racial group of physicians(Kaplan et al,1996). More research has been conducted to understand the role of physician gender in communication with patients. Methods used in these studies may provide a framework for how the role of physician ethnicity and communication style in understanding racial and ethnic disparities in healthcare might be further elucidated.

Physician gender. A large amount of research conducted in non-clinical settings has found gender differences in communication style ( Brody & Hall, 2000; Dindia & Allen, 1992; Eagly & Johnson, 1990; Hall, 1984). Indeed, the magnitude of gender differences in nonverbal expression rivals or exceeds the gender differences found for a wide range of other psychological variables (Hall, 1998). As compared with women, men have been shown to engage in less smiling and laughing, less interpersonal gazing, greater interpersonal distances and less direct body orientation, less nodding, less hand gesturing, and fewer back-channel responses (interjections such as “mm-hmm” which serve to facilitate a partner's speech), and to have more restless lower bodies, more expansive arm movements, and weaker nonverbal communication skills (in terms of judging the meanings of cues and expressing emotions accurately through nonverbal cues). Men have also been found to use less verbal empathy, to be less democratic as leaders, and to engage in less personal self-disclosure than women. Also relevant is research suggesting that women experience many emotions more frequently and more intensely than men do, and refer more to emotions in their language.

Observational studies of physician-patient communication typically have many fewer physicians than patients and a typical male-female physician ratio of 2:1. This, plus the relative recency of interest in the role of gender on the process of care, has resulted in only two dozen or so studies that have systematically compared the communication styles of male versus female physicians using audio or video recording or neutral third-party observers (Roter & Hall, 1998, Roter & Hall, 2001).

One study found that although male and female physicians did not differ in how much biomedical information they conveyed, the male physicians' talk included less psychosocial discussion. Male physicians also asked fewer questions of all sorts, engaged in less partnership-building behaviors (enlisting the patient's active participation and reducing physician dominance), produced less positively toned talk and less talk with emotional content, used less positive nonverbal behavior (e.g., smiling and nodding), and had overall shorter visits than female physicians. . Consistent with these direct observational effects, male physicians report liking their patients less than female physicians report (Hall et al., 1993) and hold less patient-centered values than female physicians (where a patient-centered response is the belief that the patient's expectations, feelings, and life circumstances are critical elements in the treatment process) (Krupat, Rosenkranz, Yeager, Barnard, Putnam, & Inui, 2000).

Based on a recent meta-analysis of the literature, the effect sizes for these gender differences are often small. However, they could have an important impact when generalized over many medical visits and many patients (Roter, Hall & Aoki, 2001,manuscript under review). To the extent that male physicians' behavior and attitudes are less patient-centered than those of female physicians, there may be implications for overall quality of care and health outcomes. Considering that the gender differences among physicians closely mirror those found in the general population, it is likely that female physicians will have fewer barriers to overcome when learning to apply the biopsychosocial model in medical practice to reduce ethnic disparities in patient-physician communication.

Physician social class (parental socioeconomic status). Medicine is practiced largely by members of the middle class and reflects middle- class ethics in terms of hard work, delayed gratification, economic independence, and autonomy (Mechanic, 1974). Medicine is also a vehicle for social mobility, but only for those who have demonstrated mastery of middle-class values through academic performance (Kurtz & Chalfant, 1991).

One effect of social class origin on the way physicians relate to patients is in terms of class-based communication styles. Several studies have demonstrated sociolinguistic differences among members of varying social classes. Reviewed by Waitzkin and Waterman (1974), the evidence suggests that there are indeed social class differences in linguistic skills. Most prominent is a tendency for middle-class subjects to be verbally explicit, while working-class subjects tend to communicate more implicitly through nonverbal signals. While most consideration of the consequences of these linguistic differences has been in terms of patients' communication, it is also possible that social-class background relates to differences in physicians' communication. These differences might act to enhance communication between physicians of lower social class origin with patients of similar social class backgrounds, or to impede the ability of physicians from poorer backgrounds to communicate with patients of higher social classes(Waitzkin & Waterman, 1974).

While not well studied, this issue has been explored. Physicians' social class background, as measured by their fathers' occupations and the physician's style of communication, was studied in audiotapes of 34 doctors in 336 medical visits (Waitzkin, 1985). When compared with doctors from upper- or upper middle-class backgrounds, physicians from working-class backgrounds tended to spend more time informing their patients, giving more explanations, and providing responses that were at the same technical level as the questions asked. The study concluded that "Orientation to verbal behavior may be a class-linked phenomenon that affects doctors as well as patients. Thus, doctors from working-class backgrounds may differ in their verbal behavior from doctors who come from a higher class position." (Waitzkin, 1985, p. 92).

Another example of the evidence of how social background affects physicians' styles of practice is found in Hollingshead and Redlich's (1958)classic study of psychiatrists in New Haven, Connecticut. The social background of the 30 psychiatrists in the study was strongly associated with how they related to patients, as well as their therapeutic orientation. Therapeutic orientation was found to fall within two distinct approaches to the treatment of patients. The first approach is analytic and psychological in orientation, with an emphasis on patient insight and as little physician directiveness as possible. These psychiatrists were almost passive in relation to management of their patients and almost never performed physical or neurologic examinations. The second approach is much more active and biomedical in nature. These psychiatrists were very directive in their therapy, often combining suggestions and advice with medical procedures, drugs, and neurologic and physical tests.

There were marked differences in the social and cultural backgrounds of the psychiatrists in these two treatment approach groups. As a group, the analytic psychiatrists had moved upward much farther in the class structure than the directive group. Almost three-quarters of the analytic group, compared with 42 percent of the directive group, moved upward one or more classes from the positions occupied by their fathers (Hollingshead & Redlich, 1958). The investigators also found that the number of generations the psychiatrist's family had been in the United States was linked to his or her theoretical orientation. Only 8 percent of the analytically oriented group were from “old American stock,” whereas 44 percent of the directive group were from that background. In contrast, 58 percent of the analytic group were first- and second- generation Americans, compared with 38 percent of the directive group.

Hollingshead and Redlich speculate that the analytic psychiatrists "like all phenomenal upward mobile persons, those who have achieved their present class positions largely through their own efforts and abilities have passed through a social, possibly also psychological, transformation," (Hollingshead & Redlich, 1958, p. 165) which accounts for their practice style. While not specifically studied, it is interesting to speculate that the psychiatrist orientations described by Hollingshead and Redlich may also apply to primary care physicians and their tendency to relate to patients in a more or less directive manner.

Relevant to the point are the findings from a large survey of physicians (Haug & Lavin, 1983), which found that those who rose to the middle class reported greater attitudinal acceptance and behavioral accommodation to consumerist-type patient challenges than those who originally came from upper- and upper-middle-class backgrounds. This may reflect a more directive and "take charge" orientation of physicians from higher social class origins than those who are upwardly mobile. Haug and Lavin (1983) note that these findings are contrary to the theory that the upwardly mobile are more conforming to traditional norms. It is difficult to predict, based upon available evidence, what role physician social class might play in understanding racial and ethnic disparities in patient-physician communication.

The Role and Impact of Patient Sociodemographics on Medical Communication

In this section we will explore the extent to which the literature presents evidence of how patient characteristics, such as race and ethnicity, gender, social class, literacy, health status, and age cohort affect patient-provider communication.

Patient race and ethnicity. Ethnic origin and cultural background contribute not only to the definition of what symptoms are noteworthy, but are also responsible for how symptoms will be presented to the physician. Studies have found that physicians deliver less information, less supportive talk, and less proficient clinical performance to black and Hispanic patients and patients of lower economic class than they do to more advantaged patients, even in the same care settings (Bartlett, Grayson, Barker, Levine, Golden, & Libber, 1984; A. Epstein, Taylor, & Sewage, 1985; Hooper, Comstock, Goodwin, & Goodwin, 1982; Ross, Mirowsky, & Duff, 1982; Waitzkin, 1985; Wasserman, Inui, Barriatua, Carter, & Lippincott, 1984). Various interpretations are possible. One is that physicians perform more poorly with such patients because they devalue them and their needs. Another is that the poor performance stems from stereotypes about the expectations, capacities, and desires of such patients. Still another is that due to cultural norms or lack of confidence, such patients do not request or demand a high level of performance from their physicians (which would, of course, confirm whatever stereotypes the physicians may already have). All of these possibilities could be relevant, for example, to the finding that Asian-American patients in Hawaii report a lower degree of participation in their medical visits than mainland Americans do (Young & Klingle, 1996). An additional example is provided by preliminary data from our ongoing study of primary care physicians and patients in the Baltimore-Washington D.C. metropolitan area showing that African-American patients experience shorter, more physician verbally dominated, and less patient-centered visits than white patients (Cooper, Roter, Ford, Steinwachs, and Powe published abstract, 2000).

Other studies of the role of patient ethnicity in the medical dialogue have been conducted in Eurocentric groups. These studies provide a basis upon which studies may be designed and analyzed for patients belonging to traditionally underserved ethnic minority groups. One classic study of health and ethnicity (Zola, 1963) found that among patients seeking medical care from several different outpatient clinics, those of Italian rather than Irish or Anglo-Saxon descent were much more likely to be labeled as having "psychiatric problems" by their physicians, despite the fact that there was no objective evidence that these problems were more frequent among this particular group. For instance, when the doctors could not identify any specific disease to explain the patient's symptoms-- which happened equally often in each of the ethnic groups-- Italians almost always had their symptoms attributed to psychological problems; this almost never happened in the case of the Anglo Saxons and Irish.

Differences were evident, however, in how the Italians presented their chief complaints. Italians reported more pain, more symptoms overall and in more bodily locations, and more consequent dysfunction, including interference with their social and personal relations. From these findings, the investigator (Zola) speculates that the Italians and Irish have ways of communicating illness that reflect different ways of handling problems within the culture itself. The Italians tend toward drama and exaggeration as a means of dissipating and coping with anxiety, whereas the Irish have a tradition in which control and denial are foremost (Barzini, 1965). This became evident in the very different ways these patients presented their symptoms to their doctors. Similar findings were reported by Zborowski (1952) in describing ethnic variations in response to pain. Anglo-Saxon patients viewed pain in an "objective" and rather unemotional way, the Irish often denied pain, and Italian and Jewish patients were highly emotional and exaggerated in their pain expression. Moreover, the Italian patients sought immediate relief from pain and were satisfied as soon as the pain ceased, but the Jewish patients were more concerned about the significance of their pain for future health and resisted pain medication for fear that it would mask a significant symptom. It is also important to note that appropriate treatment was tied to the way in which patients presented their pain. For example, painkillers would be effective for the Italian patients, but not for the Jewish patients until reassurance about future health was also provided. Only a physician sensitive to these distinctions could appropriately recognize these needs.

A follow-up study, using the same clinics as in the Zborowski study some 20 years later (Koopman, Eisenthal, & Stoeckle, 1984) found similar differences in pain reporting between Anglos and Italians. However, the effects of culture were most evident with patients over 60 years of age. Sex was also found to be important in this study; pain was most likely to be reported by older female Italians and least likely to be reported by older male Anglos. For younger patients, now second and third generation in this country, the process of acculturation had diminished the ethnic effects.

Given the correlation between social class and ethnicity in our society, it is not surprising that doctors' treatment of patients in different ethnic groups tends to parallel that for different social classes. Whites have been shown to receive care that is of higher technical and interpersonal quality than blacks or Hispanics receive, as well as to receive more positive talk and more information, even within the same medical practices (Hall, Roter, & Katz, 1988; Tuckett, Boulton, Olson, & Williams, 1985). One study found that blacks received fewer recommendations for open-heart surgery, although they had equal clinical need; of all patients who received such a recommendation, blacks had surgery less often (Maynard, Fisher, Passamani, & Pullum, 1986).

Another one of the few communication studies to address directly the issue of ethnicity found that physicians demonstrated better questioning and facilitating skills and more empathy skills when with Anglo-American as compared to Spanish-American patients (Hooper et al., 1982). The investigators suggest that poorer performance is particularly evident in communication skills requiring listening.

We believe that negative stereotypes of disadvantaged social groups affect the way doctors interact with these patients. We also believe this is unintentional and that doctors are only dimly aware of differences in their behavior, if at all. Like most people, doctors probably attribute any differences they do notice in their own behaviors to the character, aptitude, or needs of the other (in this case, the lower-class or ethnic minority patient). When one's own behavior can be construed as negative, one is particularly inclined to blame it on the other person. Attributing one's own behavior to causes outside of oneself would greatly decrease the likelihood that doctors recognize a connection between their own attitudes and behavior (Roter & Hall, 1992).

Patient gender. Among patients with chronic disease, females are more likely to prefer an active role in medical decision-making than males (Arora & McHorney, 2000). Indeed, this preference appears to be borne out in practice, as male patients report that they experience less opportunity for decision-making in their encounters with their physicians than female patients report (Kaplan, Gandek, Greenfield, Rogers, & Ware, 1995). In that study, patient participation in decision making was particularly low when male patients interacted with male physicians, a finding consistent with the finding that in male patient-male physician interactions the contribution of the patient relative to the physician is the least of all gender combinations (Hall, Irish, Roter, Ehrlich, & Miller, 1994b).

Waitzkin (1985) found that female patients were given more information than male patients, and that the information was given in a more comprehensible manner. The same dataset also revealed that the greater amount of information directed toward women was largely in response to women's tendency to ask more questions in general and to ask more questions following the doctor's explanation (Wallen, Waitzkin, & Stoeckle, 1979). Similar conclusions were reached in an English study by Pendleton and Bochner (1980), who found that female patients were given more information than males and that this information was in response to the women's more frequent questions. These findings are consistent with those relating to patient activation described above.

Investigators have also found that female patients receive more positive talk and more attempts to include them in discussion than males. In one study, physicians were more likely to express “tension release” (mainly laughter) with female patients and to ask them more for their feelings (M. Stewart, 1983).

Patient age. Patient age is associated with both the frequency of medical contacts and the communication dynamic of these visits. Older patients are plagued by multiple and complicated medical problems and are often required to make difficult decisions regarding the management of debilitating conditions. Moreover, it is the elderly who are most often required to confront fundamental choices at the end of life. While faced with these challenges, the elderly may be at a special disadvantage in fully understanding the complex choices they are asked to make. The elderly typically demonstrate lower levels of literacy and have had less exposure to formal education than younger birth-cohorts (Gazmararian et al., 1999).

Particularly relevant to these decision-making demands is the fact that older patients appear to experience medical visits during which they are more passive and less actively engaged in the treatment decision-making process. The Medical Outcomes Study (MOS) (Kaplan et al., 1995), based on surveys of over 8,000 patients sampled from the practices of 344 physicians, found that patients aged 75 and over reported significantly less participatory visits with their doctors than all but the youngest age cohorts of patients (those younger than 30 years). Interestingly, the most participatory visits were evident in the scores of only slightly younger patient groups including those aged 65 to 74, and the middle-aged group ranging from 45 to 64 years. In addition to age, ethnic minority status, poor health status and lower educational achievement were associated with lower reports of participatory visits. Thus, older ethnic minority patients may be at fourfold risk for low levels of participatory engagement with their doctors. All of the above factors may act alone or in concert to diminish the likelihood that full patient-physician partnerships will develop.

Older patient visits are also distinguished from those of younger adults by the presence of a visit companion; estimates of the percentages of all visits that include a companion range between 20% and 57% (Prohaska & Glasser, 1996). There are only a few empirical studies of the effect of companions on the dynamics of exchanges in discussions, but their presence appears to change communication patterns. Greene et al. (1987) found that when a companion was present, older patients raised fewer topics, were less responsive to topics they did raise, and were less assertive and expressive. Moreover, patients were sometimes excluded completely from the conversation when a companion was present. Additional communication difficulties have been identified, including a tendency for a family member to take on the information-giving role in the visit, sometimes contradicting the patient or disclosing information the patient had not wanted revealed (Hasselkus, 1994). It appears that the content, tone, and nature of the medical discussion may be shaped by the roles adopted by the patient companion, and these may range from advocate and supporter to antagonist (Adelman, Greene, & Charon, 1987; Greene, Adelman, Friedman, & Charon, 1994).

We know little about moderating effects such as the visit companion's ethnicity and culture, age cohort, and the nature of familial relationships on communication dynamics. An intriguing question is the effect on communication when “baby boomers” serve as companions in the medical visits of their aging parents, compared with spouses or contemporaries acting as visit companions. We might speculate that these adult children bring a assertiveness to their medical encounters that can dominate the visit and perhaps contribute to a verbal withdrawal by the patient from the medical dialogue (as described by Greene et al, 1987, above); alternatively, the presence of a consumerist companion may spur assertive behavior on the part of some patients. These dynamics are made all the more complicated when consideration is given to the impact of changing social norms for communication across racial groups, particularly for African-American birth cohorts (Satcher, 1973).

Patient social class . The effect of social class on patients' presentation of themselves and their problems also has relevance for the medical treatment patients receive. Doctors talk more with patients who are higher in social class. This has been found in Florida, Massachu setts, California, England, and Scotland (Roter & Hall 1992). How does such treatment affect the patient and what accounts for such findings? Does the higher social class patient have more to say? Or does this patient not have more to say, but instead, has the assertiveness to say it? Does the doctor give more opportunity for such a patient to talk by nonverbal indications of interest and by asking more inviting questions?

Doctors give more information to the higher-class patients, even though, when asked later, patients of different classes do not differ in how much information they say they want. In their English videotape study of 79 general consultations, Pendleton and Bochner (1980) found that patients' social class was a significant predictor of how many explanations were volunteered by doctors. Physicians spontaneously offered more explanations to patients of higher-class backgrounds during visits than to other patients. The investigators suggest that physician explanations are less likely to be volunteered to patients of lower-class backgrounds because they are perceived as less interested in information and more diffident in question-asking.

In an earlier Scottish study, Bain (1976) found that patients of lower-class backgrounds were less verbally active overall during medical visits than others. This was especially evident in areas such as patient presentation of their symptoms, question-asking, and social talk. Physicians were much more likely to give higher-class patients information regarding problem resolution and to engage in social talk with them than with lower-class patients. In further analysis of these data, Bain (1977) found that communication regarding drugs was significantly less successful with patients of lower-socioeconomic backgrounds as their recall of diagnosis, drugs prescribed, and advice given regarding how often drugs should be taken and the duration of treatment was less than other patients. , Bain's later U.S. study (1979) involving 22 physicians and a total of 556 patients, confirmed similar differences in the overall content of communications for patients from different socioeconomic groups. Patients with higher socioeconomic class backgrounds engaged in nearly 60% more talk with the physician during the visit than patients of lower socioeconomic groups.

Work by Cartwright and others appears to support the contention that patients of lower-class backgrounds appear diffident in questionasking, not because they do not wish to know about medical matters, but rather because the social distance between themselves and their physicians discourages verbal assertiveness (Cartwright, 1967). Waitzkin (1985) attributes the paucity of direct question-asking by working-class patients to their sociolinguistic culture, which tends to be less verbal than that of the middle class. Because of the tendency away from direct (verbal) communication, working-class patients may be communicating their desire for information in ways physicians are likely to miss. Doctors, like other members of the middle class, expect communication to be verbal and explicit; if patients have questions, they (physicians?) expect that they will be asked. Consequently, non-solicited information is not offered and reticence is taken as an indication of disinterest.

Waitzkin's (1985) large study in the United States found that better educated patients and patients of higher socioeconomic backgrounds received more physician time, more total explanations, and more explanations in comprehensible language than other patients. Ironically, physicians not only gave more information to higher-class patients, they also appeared to go out of their way to offer these explanations in clear, nontechnical language. Multivariate analysis of these data further demonstrated that patients' level of education was more important than social class in general in explaining information transmittal. Thus, Waitzkin concludes that the educational aspect of social class determination is a particularly strong factor in doctor-patient communication.

In a similar vein, Stewart (1983) reports that better educated patients were much more likely to receive a justification for their treatment regimens from their physicians than less educated patients. In this study, however, more information came at the price of communication, which offered emotional support. The bettereducated patients in this study received less "solidarity" from their physicians than did those patients without some university level training.

The opposite finding in regard to emotional support has been reported in several communication studies of pediatric visits wherein bettereducated parents of patients received more emotional support than less educated parents. The classic study by Korsch and associates (Korsch, Gozzi, & Francis, 1968) of pediatric encounters in an emergency walk-in clinic found that better- educated parents of patients were more likely to express their fears and hopes to the doctor and that they had a better chance of having these responded to or dealt with than less educated parents. Similarly, the pediatric study by Wasserman and associates (Wasserman, Inui, Barriatua, Carter, & Lippincott, 1983) found that better educated mothers received more reassurance, encouragement, and empathy during pediatric visits than less-educated mothers. Finally, in the most extensive observational study of pediatric practice, Ross and Duff (1982) observed indicators of performance quality, both technical and interpersonal, in over 400 pediatric visits and reported that poorly educated parents received worse care on all accounts from their physicians. Also noted in this study was that low-income families did not have as consistently negative experiences as did the children of the poorly educated. These authors concluded, as did Waitzkin (1985), that education has more significance for health experience than other socioeconomic indicators.

In sum, we can say that physicians engage in more talk overall, and especially more informative talk, with patients of higher as compared to lower social classes. Moreover, the evidence suggests that education may play a key role in the differential communication to patients of varying socioeconomic groups. The communication advantage for the better educated is especially evident in socioemotional support expressed during pediatric encounters.

It has long been known that poorer and less educated patients have trouble finding healthcare and get less of it. Now it appears that the problems of these groups are not entirely structural. They suffer poorer treatment even after they gain access to the healthcare system. The poor also have worse health, and although this has usually been assumed to stem from lifestyle factors such as stress or poor nutrition or from difficulties in getting care, the possibility must also be raised that disadvantaged patients may be sicker partly because of the way in which they and their doctors communicate.

Patient health status. The state of a patient's physical and mental health is related to both patient and physician communication (Bertakis, Callahan, Helms, Rahman, & Robbins, 1993; Hall et al., 1996). When the patient is more distressed, either physically or mentally, both the patient and the physician engage in less social conversation and make more emotionally concerned statements, engage in more psychosocial discussion, and ask more biomedical questions. Sicker patients also provide more biomedical information. The research also suggests that physicians may respond ambivalently to sicker patients; physicians report less satisfaction after visits with sicker patients, and they report liking sicker patients less than more healthy patients (Hall et al., 1993, 1996). This apparent ambivalence, in conjunction with numerous findings showing that people with worse health status are less satisfied with their care (as reviewed by Hall, Feldstein, Fretwell, Rowe, & Epstein, 1990), raises the question of whetherphysicians produce dissatisfaction in their sicker patients by displaying negative behaviors toward them. Hall, Milburn, Roter, and Daltroy (1998) used structural equation modeling to test this hypothesis and also to test whether the dissatisfaction stems directly from the sicker patient's negative outlook. In general, the direct path was supported over the physician-mediation path, with one exception: physicians' curtailing of social conversation with sicker patients accounted for some of these patients' dissatisfaction. This is unfortunate, for in curtailing this “expendable” category of interaction in the service of devoting time to more pressing medical issues, physicians may unknowingly undermine their relationships with the very patients to whom the quality of the relationship may matter most. To the extent that ethnic minority patients have poorer access to healthcare, present for care at later stages of disease, and have poorer health status, the impact of their health status may further exacerbate the communication problems they are already experiencing due to cultural or social class differences from their physicians (Hall & Roter, forthcoming, 2002).

Patient literacy. Access to healthcare is traditionally assumed to occur once the patient walks through the examining room door. Yet, providing access to care is more than simply achieving patient presence; it also involves enabling patient engagement in the process of care. While not as obvious as cost and transportation barriers, low literacy skill is a significant obstacle to full access to effective medical care (Miles & Davis,1995; Baker, Parker, & Clark, 1998). Even after adjusting for confounding sociodemographic variables such as ethnicity, gender, and age, Baker and his colleagues (1998) found that patients with the low literacy skills have poorer health, higher rates of hospitalizations, and incur higher healthcare costs than those patients with adequate literacy.

The problem of poor literacy and its significance to patient care is made more meaningful when put within the health context. The Test of Functional Health Literacy Assessment “TOFHLA,” developed by Parker and colleagues, is based on specific measures that reflect a patient's ability to perform health-related tasks that require reading and computational skills. These tasks include taking medication, keeping appointments, appropriately preparing for tests and procedures, and giving adequate informed consent(Parker et al., 1995). Using the TOFHLA measure, Baker and colleagues have demonstrated that fully one-third of patients admitted to their inner-city hospital were functionally illiterate, while another 13 percent demonstrated marginally functional health literacy(Baker et al., 1998). These findings are consistent with prior research reporting inadequate or marginal functional health literacy to be 35 percent among English-speaking patients and 62 percent among Spanish-speaking patients seeking care at public inner-city hospitals. The prevalence of low literacy among elderly patients (more than 60 years of age) was greater than 80 percent for both English- and Spanish- speaking patients (Williams et al., 1995).

With so high a prevalence of low literacy skills among the general population, physicians are likely to encounter patients with limited literacy skills. Nevertheless, physicians are unlikely to recognize patients with poor literacy skills or appreciate the negative consequences of low literacy on patient outcomes (Weiss & Coyne, 1997). Common misconceptions surrounding illiteracy may hamper such recognition. In their comprehensive overview of challenges in teaching patients with low literacy skills, Doak et al. ote several widespread myths associated with illiteracy: 1) that people with low literacy skills are intellectually impaired and slow learners; 2) that most adults with low literacy skills are poor, immigrants, and minorities; 3) that total years of schooling is a good measure of literacy levels; and 4) that people will tell you if they can not read or they will get help when they need it.(Doak et al, 1996)

In fact, most people with low literacy skills have an average IQ and function reasonably well by compensating in other ways for lack of reading skills (Doak et al., 1996; Jackson et al., 1991). The circumstances underlying low literacy are varied. Limited educational opportunity accounts for many of the poor literacy skills evident among the elderly and new immigrants, while undetected or inadequately addressed learning disabilities may account for a large proportion of the younger population who have poor literacy skills despite exposure to formal education(Kirsch et al., 1993). While low literacy is more frequent among persons of lower socioeconomic status, the poorly educated, the elderly, and American-born ethnic minorities, immigrants, and the disabled, it is not limited to these groups. According to the U.S. Department of Education's National Adult Literacy Survey (NALS), in absolute terms, the majority of the low-literate population are white native-born Americans (Kirsch et al., 1993). Furthermore, more than 20 percent of adults tested in the NALS and 18 percent of patients in the study by Baker et al who had demonstrated the lowest levels of literacy had earned a high school diploma. As noted by Baker et al., years of schoolingreflect education completed, not skills attained, and is a rather insensitive measure of literacy (Baker et al., 1998). Indeed, adults typically read 3-5 grade levels lower than years of formal education completed (Jackson et al., 1991; Doak and Doak, 1980).

Commonly held expectations about reading ability and social approbation that often follow discovery of literacy limitations may serve to inhibit disclosure, silence patients, and further discourage efforts to seek information or request assistance. Parikh and colleagues report that one in three patients characterized by the TOFHLA as having low functional health literacy did not acknowledge difficulty reading or understanding what they read.Among those patients who admitted having trouble reading, 40% acknowledged feeling shame, and more than half of these patients had never told their spouses or children about their difficulties reading (Parikh et al., 1996).

Poor literacy may be a marker for an array of communication and information processing problems that go far beyond reading ability. Communication difficulties faced by patients may be due in part to differences in vocabulary, but may also be attributed to differences in the structure and complexity of speech used by literate and low-literate populations(Roter, Rudd, & Comings, 1998). Some insight into the explanatory mechanisms linking literacy to health outcomes is provided by LeVine and colleagues in several studies conducted in developing countries (LeVine et al., 1994). These investigators have found a relationship between literacy skills level and comprehension of oral communication. The authors suggest that literacy builds a cognitive process that facilitates comprehension of formal spoken language, such as that commonly included in health messages. Even further, Dexter et al. have demonstrated that poor literacy skills are linked to an individual's ability to give adequate health-related descriptions (Dexter et al.,1998). This research suggests that low literacy not only presents obvious barriers to effective patient education, but may also complicate the process of history taking and establishment of the primary complaint.

Difficulties understanding physicians' oral communication have also been noted by patients in the United States. An analysis of focus groups and individual interviews with low-literate patients revealed serious and widespread communication difficulties with their health providers(Baker et al., 1996). Patients complained that they felt they were neither listened to nor adequately informed about their medical problems and treatments in ways they could understand. Despite this frustration, few patients asked questions or otherwise revealed their difficulties to their providers. Indeed, few patients disclosed their poor literacy to providers at all. Though some patients concealed this information out of embarrassment, others simply did not think this was something the physician would be interested in knowing.

While better-designed, learner-centered materials are undoubtedly helpful to patients, there is little offered in the literature to help physicians better communicate with their low-literate patients. Consequently, it is not surprising that when patients with poor literacy skills are recognized, few physicians feel competent to adequately respond to their needs(Miles & Davis, 1995; Weiss & Coyne,1997).

The Consequence of Race Concordance on Patient Reports of Physician Participatory Decision-Making (PDM) Style and Other Aspects of Communication

In our telephone survey study of 1816 adult managed care enrollees attending primary care practices in a large urban area, we examined the association between race or ethnic concordance and discordance on patient ratings of physicians' PDM style (Cooper-Patrick et al., 1999). The patient sample was 43% white, 45% African American, and 12% other race/ethnic groups (5% Asian, 5% Hispanic, and 2% Native American). To study the potential influence of race concordance or discordance between physicians and patients on PDM, we stratified patients according to the race/ethnicity of their physicians. We then measured the relationship between PDM style and patient race within each physician race group, adjusting for patient age, gender, education, marital status, health status, and length of the relationship. African- American patients had significantly less participatory visits with white physicians than white patients (☐= -4.3, SE=1.7, p<0.02, adjusted). Asian and Hispanic patients had less participatory visits with African-American physicians than African-American patients; however, these results were based on very small sample sizes. There were no significant racial differences in PDM scores among patients seeing Asian or Hispanic doctors. However, there were only two Hispanic physicians in the study sample; therefore, reliable conclusions regarding the participatory decision-making style of Hispanic physicians could not be drawn.

To explore the overall significance of racial and ethnic concordance in the physician-patient relationship, we conducted an analysis to assess the relationship between race/ethnic concordance between physicians and patients and PDM style. Patients in race-concordant relationships with their physicians rated their physicians as significantly more participatory than patients in race discordant relationships (☐=+2.6, SE=1.1, p<0.02, adjusted)(Cooper-Patrick et al., 1999).

In another study of the impact of racial concordance on patient ratings of care by physicians, researchers used data from a nationwide telephone survey of 2201 white, black, and Hispanic adults who reported having a regular physician (Saha et al.,1999). In this study, black respondents with black physicians were more likely than those with nonblack physicians to rate their physicians as excellent overall, at treating them with respect, explaining problems, listening, and being accessible to them. Hispanic patients with Hispanic physicians were more likely than those with non-Hispanic physicians to be very satisfied with their healthcare overall, but not significantly more likely to rate their physicians as excellent. The Hispanic respondents were primarily of Mexican and Puerto Rican descent, and the majority of them were born in the United States.

Studies have shown that ethnic minority physicians, particularly blacks and Hispanics, are more likely to provide healthcare to ethnic minority, underserved, medically indigent, and sicker populations than are their white counterparts(Moy & Bartman, 1995;Komaromy et al,1996;Xu et al,1997). Furthermore, ethnic minority patients are more likely than white patients to report having an ethnic minority physician as their regular doctor(Gray & Stoddard,1997). Nevertheless, minority patients are far more likely to receive their care from white and Asian physicians than from physicians who are African American or Hispanic since the number of physicians from these ethnic groups is so small.

Another study by Saha and colleagues, using The Commonwealth Fund 1994 National Comparative Survey of Minority Healthcare, showed that black and Hispanic respondents who had the ability to choose their physician were more likely to choose a racially or ethnically concordant physician (Saha et al, 2000). About 42% of the Hispanic respondents who chose a Hispanic physician did so because of language. Other reasons for ethnic minority patient preference for and higher ratings of care in race-concordant relationships with physicians are unclear, but potential explanations include more cultural sensitivity to the needs of these patients by race-concordant physicians and more shared cultural values, beliefs, and experiences in society between ethnic minority patients and physicians (Cooper-Patrick et al, 1999, Nickens, 1995).

Evidence of Race-Concordance Consequences for the Communication Process

In an ongoing cross-sectional study using post-visit surveys and audiotape analysis, we have examined the relationship between race concordance and actual patient and physician communication behaviors (252 adult patients 142 white, 110 African American; 31 primary care physicians-- 13 white, 18 African American). Our preliminary analysis shows significant differences in communication dynamics along several dimensions, but not exactly as expected. African- American patients appear to have shorter visits (by about 2.5 minutes) when with white physicians (13 minutes) compared to visits with African -American physicians (15 minutes). Notable were substantially longer visits between African-American physicians and their white patients (18 minutes). Same-ethnicity visits had virtually the same length of visit (15 minutes), regardless of whether the dyad was African-American or white.

Physician verbal dominance was significantly higher in visits with African-American patients than in visits with white patients (for both African-American and white doctors); the lowest levels of verbal dominance were evident in visits between African-American physicians and their white patients. Similarly, visits between African-American doctors and white patients were the most patient-centered. Visits of African-American patients were the least patient-centered, regardless of whether the physician was African-American or white. When the same survey questions used in earlier research were used to determine patient ratings of PDM, the ratings were significantly associated with audiotape measures of patient-centered interviewing (p<0.01, Cooper, Roter, Ford, Steinwachs, Powe, published abstract, 2000). However, the mean levels of PDM reported by patients are inconsistent with results of the audiotape analysis. Race-concordant dyads were rated by patients as more participatory than discordant dyads (consistent with previous work, Cooper-Patrick et al, 1999). We found lower- than- expected ratings of PDM by white patients with African-American doctors (where these visits showed the highest audiotape indicators of patient-centeredness). In contrast, higher- than- expected ratings of PDM were reported by African-American patients with African-American doctors. Patient ratings of white physicians' PDM appeared to follow the audiotape indicators of patient-centeredness more closely. PDM ratings were lowest in visits between African-American patients with white doctors (where audiotape indicators of patient-centeredness were also low) and highest in visits between white patients and white doctors (where audiotape indicators of patient-centeredness were also high).We speculate that patient expectations and stereotyping may influence patient judg ments of PDM, regardless of how the physicians actually behaved during the visit. In other work, somewhat similarly baffling results in regard to physician gender has been described (Roter & Hall, 2001). Female physicians appear to engage in more of those patient-centered behaviors (based on audiotape analysis) generally valued by patients and predictive of positive patient outcomes; nevertheless, the literature directly relating physician gender and patient satisfaction is mixed, with some studies finding higher satisfaction with female physicians and others finding the opposite. The authors speculate that whatever advantage the female physician might have by virtue of her communication performance in these visits may be offset by other patient values and prejudices that may influence patients' evaluations of female physicians, including a perceived lack of authority or expertise because of her youth and gender. The performance of doctors in race-discordant dyads may also be offset by negative patient values and prejudices, while the opposite is true in race- concordant visits. This may be made more complicated by confounding of physician youth, gender, and ethnicity. African- American physicians are generally younger than their white counterparts, and proportionately more likely to be female. (AAMC, 2000)

The pattern of interaction between African-American doctors and their white patients parallels the interaction profile of female physicians in regard to indicators of collaboration and partnership (Roter and Hall, 1998). This is apparent in two ways. First, both African-American and female physicians use more patient-centered communication, including partnership statements in their routine communication, with the effect of more actively facilitating patient participation in the medical dialogue. These statements reflect such verbal behaviors as asking for patient expectations and opinions, checking for patient understanding, and listening attentively and showing interest. Secondly, the behavior of both African-American and female physicians reflects collaboration in their tendency to be less dominant verbally during the visit, and by spending more time with their patients. These findings may be interpreted in several ways. Patients may feel freer to talk to their female and African-American physicians because they are less intimidated by them, or because patients make positive assumptions about their physicians' interest in what they have to say because they have been encouraged through the use of partnership statements, as described earlier.

African-American physicians, like female physicians, violate social expectations that ascribe high professional status to white males. The role demands of an authoritative, competent physician may be at odds with widely held social stereotypes of ethnic minority individuals as being less competent and less well trained than individuals belonging to dominant ethnic groups. As such, African-American physicians and female physicians may cause social confusion (Beach and Roter, 2000).

The Relationship Between Communication Style and Patient Satisfaction and Health Outcomes

The physician-patient relationship and its expression through the medical dialogue has been described or alluded to in the history of medicine since the time of the Greeks (Emanuel, 1961) and in the modern medical and social sciences literature for the past 50 years (Engel, 1977; Freidson, 1970; Parsons, 1951; Szasz & Hollender, 1956). George Engel's articulation of the biopsychosocial model of medical interviewing in the 1970s (Engel, 1977, 1988), later translated into a patient-centered clinical method by McWhinney (1988, 1989), has given sub stance to the prominence of a general reference to “patient-centeredness” in medical communication curricula and research (Lipkin, Putnam, & Lazare, 1995; Mead & Bower, 2000).

Patient-centered communication skills are not unlike the many other technical skills that comprise the basis of medical practice and for which proficiency is expected. Patient-centered care is defined as healthcare that is closely congruent with and responsive to patients' values, needs, and preferences(Delbanco, 1992; Gerteis, Edgman-Levitan, Daley, Delbanco, 1993). The corresponding communication behaviors include data-gathering skills (i.e., use of open-ended questions, particularly in the psychosocial domain), relationship skills (i.e., use of empathy, reassurance, support, and emotional responsiveness), partnering skills (i.e., paraphrasing, asking for patient opinion, negotiation and joint problem solving) and counseling skills (i.e., informativeness)(Lazare, Putnam, and Lipkin,1995; Roter,2000).

Visits that include more patient questions and physician information giving, a higher proportion of psychosocial and emotional statements relative to biomedical statements, and less verbal dominance by physicians are considered more patient-centered. Strong evidence links these communication behaviors to valued patient outcomes, including improvements in markers of disease control such as hemoglobin A1c and blood pressure, enhanced reports of physical and emotional health status, functioning, and pain control (Barsky et al., 1980; Greenfield et al, 1988; Kaplan et al, 1989; Giron et al;, 1998, Stewart, 1995). Specifically, visits in which the physician uses a participatory decision-making style have been associated with higher levels of patient satisfaction (Kaplan,1995). Additionally, a meta-analysis concluded that three dimensions of communication (informativeness, interpersonal sensitivity, and partnership building) were consistently associated with patient satisfaction, compliance, and recall of information(Roter, Hall, and Katz, 1988).

Implications for Physician Training and Patient Activation to Improve Patient-Physician Communication Within Culturally Diverse Populations

Many researchers have commented on the fact that physicians perform thousands of medical interviews during their career with virtually no formal training in communication skills (R. Epstein, Campbell, Cohen-Cole, McWhinney, & Smilkstein, 1993). For a long time, the assumption was that physicians naturally have adequate communication skill or that this skill inevitably develops through frequent experience. Now, medical educators agree that training is necessary, that a solid foundation of behavioral science research exists to support training programs, and that training improves the communication of physicians. Communication skills training during medical school has been shown to have effects lasting as long as five years (Maguire, Fairburn, & Fletcher, 1986).

Despite variations in the length and format of training programs, all or most of these programs focus on the principles of relationship-centered medicine (e.g., Bensing & Sluijs, 1985; Cohen-Cole, 1991; Novack, Dube, & Goldstein, 1992; Putnam, Stiles, Jacob, & James, 1988; Roter et al., 1995). The study of Novack et al. (1992) found improvements in sensitivity to psychosocial aspects of the patient's illness, the ability to relate to patients, the ability to elicit information from patients, and the ability to communicate empathy. The training program of Roter et al. (1995) emphasized physicians' ability to recognize and handle psychosocial problems. After only eight hours of training, physicians did better with their actual patients (who were audiotaped several weeks post-training) in terms of emotion handling, recognizing psychological problems, and taking a problem-solving approach-- with no increase in the overall length of the medical visit.

Smith et al. (2000) have developed a standardized training program for primary care residents that has produced very encouraging results in terms of residents' knowledge, attitudes, self-confidence, skills in interviewing patients and dealing with relationships, skills in managing and communicating with somatizing patients, and skills in educating patients. Elements in Smith's training program include setting the stage (welcoming the patient, using the patient's name, introducing self, removing barriers to communication, putting the patient at ease), agenda setting ( indicating time available, indicating own needs, obtaining list of all issues the patient wants to discuss, summarizing and finalizing the agenda), nonfocused interviewing ( appropriate use of open- and closed-ended questions, observing the patient's cues), and focused interviewing (, symptom discovery, learning personal context of symptoms, addressing emotions).

In contrast to the existence of many programs aimed at physicians and many published evaluations of such programs, relatively little research has tried to intervene with patients to improve the communication process. Classic is Roter's (1977) waiting-room intervention to increase patients' question-asking. More striking still are the experimental studies showing that a brief waiting-room intervention to increase patients' participation in the medical visit can have significant effects on health outcomes. In these studies, a research assistant reviewed the medical record with the patient, helped the patient identify decisions to be made, rehearsed negotiation skills, encouraged the patient to ask questions, reviewed obstacles such as embarrassment and intimidation, and after the visit gave the patient a copy of the medical record for that visit. In a sample of patients with diabetes, such an intervention reduced blood sugar, reduced patients' reports of functional limitations (mobility, role functions, physical activities), and improved patients' perceptions of their overall health (Greenfield, Kaplan, Ware, Yano, & Frank, 1988). Mechanisms accounting for these effects are not entirely understood because the intervention contained a number of different elements, but are likely related both to information exchange and to feelings of empowerment. In that diabetes study, experimental patients elicited more information from physicians, talked more, and were more assertive. Encouraging results have occurred in similarly designed studies using different patient populations, such as patients with ulcer disease (Kaplan, Greenfield, & Ware, 1989). Recently, an even simpler intervention, consisting of a mailed booklet designed to instruct patients in seeking, verifying, and providing information during the medical visit and to give them the opportunity to write down their concerns and questions, had significant effects on patients' information seeking and success in obtaining information. There was also a significant effect on how much information they gave to their physicians, and on self-reported adherence two weeks later (Cegala, McClure, Marinelli, & Post, 2000).

Studies show that ethnic minority, low literate, and low SES patients experience lower levels of patient-centered communication and greater verbal passivity with physicians than whites and patients with higher levels of education. While health education programs often target disenfranchised populations, those groups are hard to reach and the least able to take advantage of innovations in educational and communication technology.

Directions for Future Research

Future research to better understand the contribution of patient-provider communication to racial and ethnic disparities in care should be multifaceted, using combinations of clinical data, patient and provider surveys, audiotape and/or videotape analysis, and qualitative methods. Additionally, intervention studies that develop and test methods for improving patient-provider communication in encounters with ethnic minority patients, as well as studies that evaluate the impact of such interventions on processes and outcomes of care, are needed. Because most of the existing research focuses on African Americans, more studies including other ethnic minority groups, such as Hispanics and Asian Americans, are needed. Moreover, there is a paucity of research on the impact of using institutional resources (direct services, cultural homophilly, and institutional accommodations) to increase cultural competency. Issues identified in existing research that are of potential interest for future research include 1) the degree to which physicians' cultural competence is explained by use of patient-centered communication, 2) the degree to which other attitudes and skills are necessary in order to achieve cultural competence in patient-physician communication, and 3) identifying which patient and provider attitudes and behaviors, in addition to patient-centeredness, explain higher patient ratings of partnership and satisfaction in race-concordant patient-physician relationships. For example, preliminary results from our study of race-concordance and communication show that race-concordance, while associated with better ratings of interpersonal care from physicians, particularly for African Americans, is not directly associated with patient-centeredness as measured by audiotape analysis, implying that other attitudes and perceptions on the part of patients and physicians are in operation.

In a recent review of the literature on the provider contribution to racial and ethnic disparities in healthcare, van Ryn (2002) discusses several studies that support a central hypothesis: that provider beliefs about patients and provider behavior during patient encounters are independently influenced by patient race/ethnicity. Van Ryn goes on to suggest that a deeper understanding of automatic and social cognition processes involved in the patient-provider relationship will allow the development of better evidence-based interventions that target patient-provider communication to address racial and ethnic disparities in care. Our review of the literature supports the hypothesis put forth by van Ryn and extends it to include the importance of reciprocity in the patient-physician relationship and the impact of patient expectations and judgments of physicians, placing all of these factors within the context of other patient and physician sociodemographic factors such as age, gender and social class.

In summary, we provide the following considerations for future research. First, more indepth exploration of social cognitions and stereotyping behavior by patients and physicians is needed. Furthermore, the reciprocal nature of the patient-physician relationship is a critical factor that should be highlighted in future work. Studies of race-concordant patient-provider relationships will provide opportunities to increase our understanding of the constructs of cultural competence. Finally, inasmuch as physician ethnicity is confounded with age and gender (minority physicians are more likely to be female and younger than white physicians), future research will require disentanglement of the complex interactions among patient and physician age, gender, and ethnicity and their impact on patient-physician communication.

The Role of Cross-Cultural Training for Healthcare Professionals

Researchers and medical educators have developed models for cross-cultural training of healthcare professionals (Sue, Zane, and Young, 1994; Gardenschwartz & Rowe, 1998;Carrillo, Green, and Betancourt, 1999; Loudon et al., 1999) that typically include the following objectives: 1) increase learners' awareness or mindfulness of the impact of culture and cultural differences on their own and other's behaviors during the process of communication between themselves and dissimilar others; 2) increase learners' knowledge of critical concepts useful in intercultural adjustment, including knowledge that others find essential for success in their own culture; 3) challenge learners' emotional states engendered by intercultural interactions and facilitate their progression through developmental stages of intercultural sensitivity (i.e., denial, defense, minimization, acceptance, adaptation, and integration); and

4) develop culture-general (ability to manage stress, tolerate ambiguity, establish realistic expectations, acquire “antennae” for cultural differences in interpersonal interactions, develop empathy with emotions expressed by dissimilar others) and culture-specific skills (those that are unique to a given people or context) (Bennett 1986; Brislin and Yoshida, 1994). These objectives have face validity; however, there is scant evidence that relates any of these objectives of cultural competence training to having an impact upon patient outcomes.

An alternative description of the cultural competence process incorporates many dimensions of patient-centeredness. Campinha-Bacote (1999) describes cultural awareness, cultural knowledge, cultural skills, cultural encounters, and cultural desire as constructs of cultural competence. Cultural awareness is the deliberate, cognitive process in which healthcare providers become appreciative and sensitive to the values, beliefs, lifestyles, practices, and problem-solving strategies of clients' cultures. The cultural awareness process involves examination of one's own prejudices and biases toward other cultures and in-depth exploration of one's own cultural background.

Cultural knowledge is the process of seeking and obtaining a sound educational foundation concerning the worldviews of various cultures. The goal of cultural knowledge is to understand clients' worldviews, or the way individuals or groups of people view the universe to form values about their lives and the world around them. Additionally, the process of cultural knowledge involves the process of obtaining knowledge regarding specific physical, biological, and physiologic variations among ethnic groups.

Cultural skill is the ability to collect relevant cultural data regarding clients' health histories and presenting problems, as well as accurately performing a culturally specific cultural physical assessment. The literature offers several assessment tools that healthcare providers can use when conducting cultural assessments (Berlin & Fowkes, 1983, Kleinman, Eisenberg, & Good, 1978, Buchwald et al, 1994). These tools essentially serve as reminders for physicians to use patient-centered communication skills when they are caring for patients from diverse cultural backgrounds.

Cultural encounter is the process that encourages healthcare providers to engage directly in cross-cultural interactions with clients from culturally diverse backgrounds. The larger the number of cultural encounters a healthcare provider has with individuals from specific ethnic groups, the more likely he/she will gain an appreciation for intra-ethnic variations and the less likely he/she will be to engage in stereotyping behavior that might develop as a result of academic knowledge without direct experience. Cultural desire is the motivation and genuine desire of health providers to engage in the process of cultural competence. This is based on true caring (empathy), curiosity or interest, and respect regarding clients whose cultures differ from one's own. We believe cultural skills and cultural desire include most of the dimensions of patient-centeredness that have been shown to have positive and significant effects on patient outcomes.

Conclusions

To the extent that cultural skills and cultural desire are synonymous with patient centeredness, more broadly defined communication skills programs that prepare healthcare providers to deliver high quality interpersonal and technical healthcare to an ethnically diverse population are promising as a strategy to reduce ethnic disparities in healthcare, particularly when these programs are consistent with patient empowerment strategies. Patient empowerment strategies increase patients' ability to 1) fully participate in the medical interview, 2) negotiate treatment plans by engaging in joint problem-solving and collaborative treatment decision-making with physicians, and 3) gain confidence and competence in illness self-management, including adherence to treatment and management recommendations (Roter, 2000). Because strong evidence links these communication strategies to improved patient outcomes, intervention studies that target ethnic minority patients and incorporate communication skills training for physicians and empowerment strategies for patients should be conducted. These studies should include rigorous measurement of healthcare processes and outcomes.

The centrality of patient voice is all the more critical in recognition of the cultural diversity that often characterizes vulnerable populations and distinguishes them from the majority culture. Ethnic minorities, the elderly, the poor, and the non-functionally literate are victims of disparities in health and the receipt of health services. (Office of Minority Health, Washington, DC, 2000). This disparity appears to be growing and its elimination constitutes a priority area for the future direction of both provider and patient education efforts.

The progressive transformations of the patient-provider relationship will depend on its embracing a broad empowerment agenda (Roter et al., 2001). This will challenge the profession on many levels and test our commitment not only to serve patients' needs, but to empower patients themselves to define those needs. The broadening definition of quality in medical care of the past decade has called for systematic efforts to incorporate the patient's perspective in defining their own medical goals and actively participating in management and treatment considerations, and to include patients in the judgment of their own functioning and well-being(Maloney & Paul).

The challenge for transforming the practice of medicine to more effectively meet the needs of ethnically diverse patients will include the generation of racial and ethnic-neutral social norms regarding patient expectations and judgments of physician conduct, as well as the establishment of medical practice norms that value communication skills, interpersonal sensitivity, and cultural competence. Increasing diversity in the physician workforce will help contribute to a societal norm that does not inherently define "doctor" in gender or race-linked terms, but this will not be sufficient to transform medical practice. Until we have more evidence of the impact of institutional resources on improving cultural competency, physician training in interpersonal skill, emphasizing those aspects of communication identified with documented benefits on patient health (e.g., patient-centeredness), coupled with patient activation and empowerment strategies, are a promising as the vehicle by which the ethnic minority patient's voice and perspective will be enhanced. Interventions incorporating these strategies will likely improve quality of care and outcomes for all patients, but those at highest risk -- the poor, elderly, chronically ill, low literate, and ethnic minority patients, may stand to benefit the most from these efforts.