Medical Decisions Under Time Pressure with Limited Information

In the process of healthcare, doctors and other healthcare providers often must reach judgments about patients' conditions and make decisions about treatment without complete and accurate information. Moreover, they frequently must do so under severe time pressure and resource constraints. These conditions contribute to clinical uncertainty, as providers must weigh a vast array of information, presented both by the patient and from diagnostic test data. This uncertainty opens the possibility that medical decisions and the course of treatment will reflect subjective variability and preferences of the physician (Eisenberg, 1986; Wennberg, 1999). Under conditions of time pressure, problem complexity, and high cognitive demand, physicians' attitudes may therefore shape their interpretation of this information and their expectations for treatment, such as the likelihood of patient compliance.

To add to this uncertainty, as clinicians and their patients work together, both parties are involved in highly complex processes of decision-making, requiring the acquisition of a wide array of diverse information and the weighting of these data on various dimensions of salience. The assembly and use of these data are affected by many influences, including various heuristics that introduce significant problems for recall and weighting. In conditions such as these, it may be assumed that cognitive shortcuts have significant value to any decision-maker. Physicians, in fact, are commonly trained to rely on gestalts that functionally resemble the application of “prototypic” or stereotypic constellations. That is, physicians use clusters of information in making diagnostic and other complex judgments that must be arrived at without the luxury of the time and other resources to collect all the information that might be relevant. These conditions of time pressure and resource constraints are common to many clinical encounters, and map closely onto those identified as producing negative outcomes due to lack of information, to stereotypes, and to prejudice (van Ryn and Burke, 2000; van Ryn, 2002).

Patients may also hold stereotypes of clinicians that would come into play under these conditions of stress and demand for rapid and complex cognition. These stereotypes may paint the physician as an arrogant clinician, or as “the white man who experiments on minority patients,” or as a person who cannot be trusted to provide the whole truth. Even if the parties would, upon direct inquiry, deny the reality of such stereotypes in the particular circumstance, they may still unconsciously act from these perspectives in a pressured situation. The following sections present a more detailed discussion of the mechanisms by which disparities can arise in the clinical encounter. We explain how the mechanisms work, and consider the evidence in support of the empirical importance of each mechanism.

Healthcare Provider Prejudice or Bias

Prejudice is defined in psychology as an unjustified negative attitude based on a person's group membership (Dovidio et al., 1996). Prejudice, when held explicitly, may become part of a “reasoned” and normative pattern of behavior that becomes discriminatory. While it is reasonable to assume that the vast majority of healthcare providers find prejudice morally abhorrent and at odds with their professional values, healthcare providers, like other members of society, may not recognize manifestations of prejudice in their own behavior. Socially conditioned implicit prejudice may be manifested in healthcare providers' nonverbal behaviors reflecting anxiety (e.g., increased rate of blinking), aversion (e.g., reduced eye contact) or avoidance (e.g., more closed postures) when interacting with minority rather than white patients.

Empirical support for the presence of biased or prejudicial attitudes among healthcare providers is limited but growing. Some research suggests that differences in care may result from conscious or unconscious biases on the part of physicians and other healthcare providers. Schulman et al. (1999), for example, assessed physicians' recommendations for management of chest pain after they viewed vignettes of “patients” (actually actors) who complained of symptoms of coronary artery disease. “Patients” varied only in race (black or white), sex, age (55 or 70 years), level of coronary risk, and the results of an exercise stress test. As originally reported in the published findings, Schulman et al. found that physicians were less likely to recommend cardiac catheterization procedures for women (odds ratio = 0.6, suggesting that they were 40% less likely to be recommended for catheterization) and African Americans (odds ratio = 0.6, again suggesting that this group was 40% less likely to referred for catheterization) than for whites and men.

These results as reported, however, overstated the likelihood of referral for African Americans and women relative to whites and men. In a rebuttal to Schulman et al., Schwartz, Woloshin, and Welch (1999) demonstrated that had the study authors calculated the relative chance of referral using risk ratios, rather than odds ratios, the probability of African Americans being referred for cardiac catheterization was only 7% lower than for whites. In addition, Schwartz et al. demonstrated that significantly lower rates of referral were found only in the case of African-American women, whose rate of referral was approximately 12% less than that for white men, white women, and African-American men. There were no significant differences in rates of referral among the latter three groups, suggesting that a more accurate interpretation of the data would be that the effect of race on physician's referral patterns is modified by gender. Furthermore, Schwartz et al. contend that referral rates for catheterization alone do not constitute a “gold standard” of care; to the contrary, these authors assert, the assumption that “more testing” – i.e., catheterization for all patients who present with cardiac symptoms – represents better care is unfounded. Less testing, in some instances, may result in more appropriate care (Schwartz et al., 1999).

In a reply to Schwartz et al. (1999), Schulman, Berlin, and Escarce (1999) agree that calculation of risk ratios would have been more appropriate as a means of assessing differences in referral rates. They note, nonetheless, that the findings of the study are consistent with the hypothesis that clinical decision-making may be influenced by physicians' conscious or subconscious perceptions on the basis of patients' race and gender, rather than on objective data. And while catheterization may not be appropriate for all patients even given similar objective preliminary test results, Schulman and colleagues, referring to the robust findings of studies that indicate lower rates of referral among blacks for catheterization, write, “we doubt that the lower utilization rates observed consistently among black patients reflect an effort to provide more appropriate care to these patients” (Schulman, Berlin, and Escarce, 1999, p. 286).

In another experimental design, Abreu (1999) assessed whether conscious or non-conscious stereotypes would influence the clinical impressions of mental health professionals. Arbeu “primed” these clinicians with either African-American stereotypes (e.g., “Negroes,” “blacks,” “blues,” “rhythm”) or neutral words (e.g., “water,” “then,” “about,” “things”) flashed on a computer screen for 80 milliseconds. Clinicians were then asked to evaluate the same hypothetical patient on a number of dimensions, including general impressions as well as clinical features. Abreu found that therapists primed with stereotype-laden words rated the patient significantly less favorably on hostility-related attributes than therapists exposed to neutral words, demonstrating that “therapists can be affected by African-American stereotypes in ways that produce negative or positive first impressions, depending on the nature of the attribute that is rated” (Abreu, 1999, p. 387).

Another experimental study using patient vignettes also found variations in physician recommendations when patient demographic variables were manipulated, although results were mediated by physician gender. In a study of primary care physicians' recommendations for pain management, Weisse et al. (2001) presented vignettes of patients suffering from identical symptoms of kidney stone pain, lower back pain, and as a control condition, sinusitis. Nearly 80% of the physician sample was white, while 15% were Asian American or Pacific Islander. In each case, only the race (African American or white) and gender of the “patient” was manipulated. The authors found that male physicians prescribed higher doses of hydrocodone for white “patients” than black “patients” suffering from back pain and renal colic, while female physicians prescribed higher doses of analgesic for black “patients” than white “patients.” In both cases, findings were robust: male physicians prescribed twice as much hydrocodone to white patients than black patients, while female physicians prescribed the reverse. No other patient-physician race and gender interactions were observed. These findings, the authors suggest, imply that male and female physicians may react differently to gender and/or racial cues. While few other studies have replicated this finding, the study also implies that healthcare providers' perceptions of and attitudes towards patients are potentially influenced by a range of factors, and illustrates the complexity of disentangling the effects of race, ethnicity, and gender as they influence patient-provider interaction. More research is needed to better understand these processes, and to specifically assess how physicians' race, ethnicity, or gender may influence their attitudes toward and perceptions of patients.

Another experimental study, using first- and second-year medical students as subjects, assessed whether the race and gender of hypothetical patients influenced students' perceptions of presenting symptoms. Rathore et al. (2000) randomly assigned 164 medical students to view a video of either a black female or white male actor who presented with the same symptoms of angina. Students were then asked to rate the patients' health status, based on their assessment of how the patients' presenting symptoms would affect their quality of life. They were also asked to provide a diagnosis of “definite” or “probable” angina. The authors found that students were more likely to provide a diagnosis of “definite” angina for the white male patient than the black female patient, but rated the health status of the black female patient as lower than that of the white male. Thus, these subjects assessed the while male patient's cardiac symptoms to be more severe, yet perceived the black female patient's quality of life to be lower, despite objectively similar presentations from the two “patients.” Minority students, however, did not rate the health status of the black female patient as significantly different than that of the white patient. When examined by students' gender, Rathore et al. found that the male students tended to rate the black female's health status as lower than the white male, while female students did not rate the two patients' heath status differently (Rathore et al., 2000).

In a study conducted in a clinical setting, Finucane and Carrese (1990) assessed when and how patients' race was referenced during house staff case presentations. In this study, the chief medical resident surreptitiously recorded oral case presentations during a 2-month period, and assessed, using a priori criteria, whether and how often the patient's race was mentioned, and whether potentially “unflattering characteristics” (e.g., low intelligence, uncooperativeness, unkemptness) were also noted. Race was noted in the vast majority (16 of 18) of cases involving black patients, but only in about half (19 of 36) of cases involving white patients. Among patients to whom house staff ascribed unfavorable characteristics, race was mentioned in 10 of 10 cases involving black patients, but in only 4 of 9 cases involving white patients. Findings of this study must be interpreted with caution, however, as the study suffers from a very small sample size and is limited by the single study setting. In addition, the authors employed no objective means of assessing whether the unfavorable characteristics ascribed to patients resulted from a true difference in the prevalence of these characteristics, rather than from racial bias.

In another study based on actual clinical encounters, van Ryn and Burke (2000) surveyed 193 physicians to assess their perceptions of 842 patients (57% white and 43% African American) following post-angiogram hospital visits. The authors asked physicians to rate their patients on a variety of personal characteristics such as intelligence, self-control, education level, pleasantness, rationality, independence, and responsibility. In addition, the authors asked physicians to rate their feelings of affiliation toward the patient and their perceptions of their patients' degree of social support, tendencies to exaggerate discomfort, likelihood of complying with medical advice, likelihood of drug or alcohol abuse, as well as other characteristics. van Ryn and Burke also surveyed patients and assessed their frailty/sickness, depressive symptoms, social assertiveness, feelings of self-efficacy, and perceived social support. These variables, along with information about physicians' age, sex, race, and medical specialty were entered into logistic regression analyses to control for the impact of these variables on physicians' assessments of patients. The results supported the authors' hypotheses that patient race and socioeconomic background do influence physicians' perceptions, even when controlling for differences in patients' socioeconomic status, personality attributes and degree of illness. African-American patients were rated as less intelligent, less educated, more likely to abuse drugs and alcohol, more likely to fail to comply with medical advice, more likely to lack social support, and less likely to participate in cardiac rehabilitation than white patients. Furthermore, African-American patients were two-thirds as likely as whites to be perceived as the kind of person the physician could see him/herself being friends with. Finally, a significant interaction of race and socioeconomic status was found, in that at low socioeconomic (SES) levels, black patients were rated as less pleasant and less rational than whites.

These studies lend support to the hypothesis that physicians' diagnostic and treatment decisions are influenced by patient race. In addition, they suggest that these influences are complex, and that both patient and provider gender may significantly influence physicians' perceptions. They do not, however, elucidate the mechanisms by which these attitudes, biases, and stereotypes may result in differences in clinical treatment, or the degree to which these attitudes might affect the outcome of patient care. It therefore remains unclear what degree of racial and ethnic disparities may be explained by this mechanism.

As noted above, there is no evidence that any significant proportion of healthcare professionals in the United States harbors overtly prejudicial attitudes. Health professionals in general are well educated and subscribe to a professional ethic that should mitigate against discrimination on the basis of race or ethnicity. How then, could a well-meaning group of healthcare professionals, working in their usual circumstances with diverse populations of patients, create a pattern of care that appears (on the now substantial weight of available scientific evidence) to be discriminatory? In other words, is it possible for physicians and other healthcare professionals to act in a racially biased manner without knowing it?

To begin to address this question, the following section offers a hypothesis about clinical uncertainty, and how it may affect healthcare providers' decision-making, and ultimately influence the care provided to minority patients.

Patient Response: Mistrust and Refusal

As noted above, racial and ethnic minority patients' responses to healthcare providers are also a potential source of disparities. Little research has been conducted on how patients may influence the clinical encounter. It is reasonable to speculate, however, that if patients convey mistrust, refuse treatment, don't adhere or comply poorly with treatment, providers may become less engaged in the treatment process, and patients are less likely to be provided alternate treatments and services. As noted in Chapter 3, some evidence suggests that patient refusal may contribute to disparities in care. For example, African-American and other minority patients may be more likely to refuse invasive procedures. This higher rate of refusal of recommended treatments may reflect patients' experiences of discrimination in other sectors or mistrust of authority. Some mistrust and refusal, however, might be a “rational” reaction to explicit discrimination, aversion, or disregard displayed by the provider. If minority patients perceive that their provider has a lower regard for them, they will be less likely to comply with treatment recommendations.

It should be noted, however, that despite ethnic minority patients' generally higher levels of mistrust of the medical and research establishment, most minority patients appear to be satis fied with and have confidence in their healthcare providers (Shi, 1999). Further, as Geiger (this volume) and others have noted, mistrust or perceived discrimination alone is unlikely to cause ethnic minority patients to reject potentially life-saving or highly recommended procedures that promise to improve health and decrease symptoms of illness. Therefore, future analyses of patient attributes that may be related to healthcare disparities must carefully consider the roots of these attitudes in historic and contemporary social and cultural forces, in and outside medical practice, that play a role in minority patients' perceptions of healthcare institutions.

In the absence of careful study as to how patients may influence the clinical encounter and contribute to disparities in healthcare, the committee is reluctant to speculate on how and to what extent such processes occur. It may be reasonable to assume, however, that patients' and providers' behavior and attitudes influence each other reciprocally and reflect the attitudes, expectations, and perceptions that each has developed in a context where race and ethnicity are often more salient than these participants recognize. In addition, it is clear that the healthcare provider, rather than the patient, is the more powerful actor in clinical encounters. Providers' expectations, beliefs, attitudes, and behaviors are therefore likely to be a more important target for intervention efforts.

Conclusion

In the previous sections, we have considered factors arising out of doctor-patient interactions that may account, at least in part, for racial and ethnic disparities in healthcare. The committee's focus has been on understanding the processes that may underlie these biases. We propose that these processes have their origins in pervasive and normal distinctions based in social categorization (stereotypes, prejudice, and uncertainties in intergroup communication) and do not necessarily involve either awareness or conscious motivations to discriminate. Thus, even highly educated and socially conscious individuals, such as doctors, are susceptible to these biases. Moreover, the types of situations that promote these biases – time pressure, incomplete information, high demand on attention and cognitive resources – are those that frequently occur in the context of doctor-patient interactions. We supported these propositions with research and illustrated their likely effects.

Beyond identifying the pervasiveness and importance of these factors in healthcare outcomes, this perspective emphasizes two other fundamental issues. First, this approach highlights the fact that disparities in healthcare services may not necessarily be a matter of “less.” Within the models of bias, with the exception of the simple prejudice mechanism, the implications of the other mechanisms may be more or less in terms of quality of services. The importance of disparities in services is that minorities may have healthcare services poorly matched to their needs. A focus on the issue of matching needs to services is a more general and pertinent framework than simply focusing on equal amounts of services.

A second implication of this perspective is that it suggests different types of policies and interventions to address disparities based on different processes. The research on healthcare disparities to date does not consistently differentiate among the various mechanisms that may operate in doctor-patient interactions and underlie the disparities. At a general level, making good choices about alleviating disparities should be based on a good idea as to what causes disparities.

In summary, the committee found no direct evidence that racism, bias, or prejudice among healthcare professionals affects the quality of care for minority patients, such as that which might be available from audit studies where “testers” from different racial or ethnic groups present in clinical settings with similar clinical complaints, histories, and symptoms to assess possible differences in the quality of their treatment. In addition, no survey data suggest that even a small minority of physicians, nurses, or other healthcare professionals harbors biases or prejudices against minorities. Both of these forms of evidence present methodologic (and in the case of paired testers, ethical and legal) challenges to investigators, making it unlikely that such evidence will be available in the near future.

In the meantime, the committee is confronted with several “streams” of evidence that, while not definitive, collectively provide a sufficient base from which to draw inferences. To summarize the evidence presented in this chapter and the previous two chapters that provider prejudice, stereotyping, and biases may influence clinical care:

1.

With increasing sophistication, several recent studies of racial and ethnic disparities in receipt of health services have controlled for possible confounding variables or other possible explanations for racial and ethnic differences in care, including patient preferences, overuse of services by whites, health insurance status, type of health system, patient income and education, severity or stage of disease, co-morbidity, hospital type, and resources. These studies generally find that disparities remain and cannot be fully explained by these variables. While this literature does not provide any measure of evidence that provider biases and stereotyping explain disparities, they do illustrate that disparities cannot be “reduced” to patients' preferences or other explanations.

2.

Racial and ethnic disparities in healthcare emerge from an historic context in which healthcare has been differentially allocated on the basis of social class, race, and ethnicity. Unfortunately, despite public laws and sentiment to the contrary, vestiges of this history remain and negatively affect the current context of healthcare delivery. And despite the considerable economic, social, and political progress of racial and ethnic minorities, evidence of racism and discrimination remain in many sectors of American life.

3.

Evidence from patient surveys indicates that racial and ethnic minority patients are far more likely than white patients to believe that discrimination is a problem in healthcare, and that they have personally experienced discriminatory treatment. Data from the focus groups conducted by the study committee suggest that minority patients may perceive both overt, as well as subtle forms of discrimination when seeking care.

4.

There is considerable evidence that even well-meaning whites who are not overtly biased and who do not believe that they are prejudiced typically demonstrate, on average, unconscious implicit negative racial attitudes and stereotypes (e.g., Dovidio, Brigham, Johnson, & Gaertner, 1996). Both implicit and explicit stereotypes significantly shape interpersonal interactions, influencing how information is recalled and guiding expectations and inferences in systematic ways. They also can produce self-fulfilling prophecies in social interactions, in which the stereotypes of the perceiver influence the interaction with others in ways that conform to stereotypical expectations (e.g., Jussim, 1991).

5.

Experimental evidence indicates that healthcare providers are influenced by patients' race or ethnicity, and possibly gender (Schulman et al., 1999; Weisse et al., 2001), or when providers are “primed” with racial stereotypes (Abreu, 1999). Preliminary evidence also suggests that female physicians may respond to racial cues differently than male physicians (Weisse et al., 2001; Rathore et al., 2000). Minority race or ethnicity is found to be associated with generally more negative evaluations or lower rates of referral for clinical services, even when “patients” present with the same clinical condition. In addition, a survey of physicians following actual clinical encounters demonstrates that physicians endorse stereotypes about their African-American patients (who were characterized as “less intelligent, less educated, more likely to abuse drugs and alcohol, more likely to fail to comply with medical advice,” and less likely “to be . . . the kind of person the physician could see him/herself being friends with”), even after controlling for patients' socioeconomic status, personality variables, and perceived social support (van Ryn and Burke, 2000).

6.

The conditions in which many medical encounters take place – characterized by time pressure, resource constraints, and high cognitive demand – have been identified in the social psychological literature as conditions that may promote stereotyping due to the need for cognitive “shortcuts” and lack of full information to adequately assess patients.

These streams of evidence lead the committee to conclude that bias, stereotyping, prejudice, and uncertainty on the part of healthcare professionals cannot be ruled out – and indeed, appear among the many patient-level, system-level, and clinical encounter-level factors to contribute to racial and ethnic disparities in healthcare.

Finding 4-1: Bias, stereotyping, prejudice, and clinical uncertainty on the part of healthcare providers may contribute to racial and ethnic disparities in healthcare. While indirect evidence from several lines of research supports this statement, a greater understanding of the prevalence and influence of these processes is needed and should be sought through research.

Indirect evidence indicates that bias, stereotyping, prejudice, and clinical uncertainty on the part of healthcare providers may be contributory factors to racial and ethnic disparities in healthcare. Prejudice may stem from conscious bias, while stereotyping and biases may be conscious or unconscious, even among the well intentioned. Ambiguities in the interpretation of clinical data, barriers to patient-provider communication, and gaps in evidence of the efficacy of clinical interventions contribute to uncertainty, and therefore may promote the activation of prejudice and stereotypes. However, few studies have attempted to assess these mechanisms, and therefore direct evidence bearing on the possible role of these factors, especially prejudice, is not yet available. The committee finds strong, but circumstantial evidence for the role of bias, stereotyping, prejudice, and clinical uncertainty from a range of sources, including studies of social cognition and “implicit” stereotyping, but urges more research to identify how and when these processes occur.

Patients' refusal or acceptance of recommendations for treatment, like other patient decisions, is the result of many influences, including information about their condition, information about treatment effectiveness and risks, trust of the clinician, preferences for treatment type and outcome, and advice of significant others. Overall, such preferences for care should be developed by patients and their families on the basis of full and accurate information presented by a healthcare provider, but the acquisition and use of such information may be influenced by the quality of patient-provider communication and interaction, patients' expectations, values and beliefs, as well as the values and beliefs of patients' communities. To the extent that minority patients are more likely than whites to refuse treatment, such behaviorally expressed preferences may be considered a source of healthcare disparities. A small number of studies suggest that racial and ethnic minorities are slightly more likely than whites to refuse treatment, but this research has yet to distinguish the sources of minority patients' higher rates of refusal (i.e., general mistrust of healthcare providers, real or perceived experiences of discrimination in healthcare settings, or patient treatment decisions based on incomplete information from providers). These sources must be better understood to fully understand the role of patient preferences in healthcare disparities.

Finding 4-2: A small number of studies suggest that racial and ethnic minority patients are more likely than white patients to refuse treatment. These studies find that differences in refusal rates are generally small and that minority patient refusal does not fully explain healthcare disparities.

A small number of studies suggest that racial and ethnic minorities are more likely to refuse treatment. These studies find that differences in refusal rates are generally small and that minority patient refusal does not fully explain healthcare disparities. However, research has yet to distinguish the sources of minority patients' higher rates of refusal (i.e., general mistrust of healthcare providers, real or perceived experiences of discrimination in healthcare settings, or patient treatment decisions based on incomplete information from providers). These sources must be better understood to fully comprehend the role of patient preferences in healthcare disparities.

Footnotes

1

A number of other explicit decision-theoretic approaches explore clinical decision-making. For other examples, see Mushlin et al. (1997) or Fendrick et al. (1995).