|Home | About | Journals | Submit | Contact Us | Français|
To describe how and why attending physicians respond to learner behaviors that indicate negative attitudes toward patients.
Inpatient general internal medicine service of a university-affiliated public hospital.
Four ward teams, each including an attending physician, a senior medicine resident, two interns, and up to three medical students.
Teams were studied using participant observation of rounds (160 hours); in-depth semistructured interviews (n = 23); a structured task involving thinking aloud (n = 4, attending physicians); and patient chart review. Codes, themes, and hypotheses were identified from transcripts and field notes, and iteratively tested by blinded within-case and cross-case comparisons.
Attending physicians identified three categories of potentially problematic behaviors: showing disrespect for patients, cutting corners, and outright hostility or rudeness. Attending physicians were rarely observed to respond to these problematic behaviors. When they did, they favored passive nonverbal gestures such as rigid posture, failing to smile, or remaining silent. Verbal responses included three techniques that avoided blaming learners: humor, referring to learners’ self-interest, and medicalizing interpersonal issues. Attending physicians did not explicitly discuss attitudes, refer to moral or professional norms, “lay down the law,” or call attention to their modeling, and rarely gave behavior-specific feedback. Reasons for not responding included lack of opportunity to observe interactions, sympathy for learner stress, and the unpleasantness, perceived ineffectiveness, and lack of professional reward for giving negative feedback.
Because of uncertainty about appropriateness and effectiveness, attending physicians were reluctant to respond to perceived disrespect, uncaring, or hostility toward patients by members of their medical team. They tended to avoid, rationalize, or medicalize these behaviors, and to respond in ways that avoided moral language, did not address underlying attitudes, and left room for face-saving reinterpretations. Although these oblique techniques are sympathetically motivated, learners in stressful clinical environments may misinterpret, undervalue, or entirely fail to notice such subtle feedback.
The essence of professional education includes transmitting not only knowledge, but also values and attitudes from one generation to the next. Training more humane physicians has been an enduring concern,1 and deans and residency directors face constant exhortation to do so.2 Yet medical students and residents grow more cynical and less compassionate over the course of their education.3, 4 Commentators agree that desirable professional characteristics include positive patient-directed attitudes such as compassion and respect.2, 5 But even medical educators deeply committed to fostering these attitudes question whether they can be taught systematically.6 If physicians are to do a better job of treating patients with compassion and respect, medical education and residency training must do better at fostering these attitudes. But to make progress in this area, we must first understand how clinical teachers presently assess, respond to, and try to shape the attitudes of medical students and residents.
Contemporary medical ethics education has concerned itself with teaching either facts 7 or moral reasoning processes,4 but not with the motivational network of values, attitudes, and feelings that underlies moral behavior.8, 9 There have been efforts to increase preclinical medical students’ moral sensitivity and empathy through literature, film, and other experiential media,10 but few such programs have been created for students or residents in the clinical training years. As the teacher with the most direct contact during those years, the attending physician has been deemed largely responsible for promoting trainees’ attitudinal development, but given little guidance as to how to do so. The conventional wisdom has remained that values are passively “caught” rather than actively “taught”: that senior students and trainees emulate the values modeled by senior clinicians,11 and that such attributes are best taught by demonstration and example in clinical contexts.12 Furthermore, the service demands of residency training restrict the time and energy available for any extracurricular intervention. So although medical educators and investigators shy away from the hard questions about whether or how attitudes can be taught, attending physicians are tacitly expected to recognize and assess attitudes and attempt to mold them.
Influencing attitude has both positive components, whereby desirable elements are encouraged or reinforced, and negative ones, encompassing efforts to modify or discourage undesirable attitudes. Attitudes like compassion and respect are easier to recognize than to define, and are often particularly recognizable in the breach. Both teachers and learners may be more likely to notice and recall responses to problematic behaviors—those that suggest a deficiency of appropriate attitude—than they would be to remember praise or positive feedback. To open a window on the teaching of attitudes, we asked: What do clinical teachers do when they believe they are witnessing deficiencies in compassion and respect? Which behaviors of trainees raise these concerns? How do attending physicians respond, and why?
We used a multimethod, qualitative, case study approach,13 to study four teams on the inpatient general internal medicine service of a university-affiliated public hospital. Teams were led by full-time faculty attending physicians with a minimum of 8 years’ experience, who were selected in order to achieve a balance of gender, age, subspecialty, and reputed teaching style. Attending physicians are referred to pseudonymously as Drs. Arthur, Baker, Cooper, and Dunn. Each team included the attending physician, a senior medicine resident, two interns, and up to three medical students.
Teams were studied intensively for 2 weeks each, using four data-collection techniques: participant observation; a structured cognitive task that involved thinking aloud; in-depth, semistructured interviews; and patient chart review. The principal investigator (JHB) followed each team as a silent participant observer,14, 15 during morning work rounds and attending rounds. Total direct observation time was approximately 40 hours with each team, with the attending physician present approximately half the time. Field notes taken during rounds were summarized immediately afterward. Attending rounds were audiotaped.
All 23 team members were interviewed by the principal investigator. Before the interview, all four attending physicians were given a 30-minute think aloud exercise. They were first given an unrelated brief written vignette with which to practice thinking aloud, a technique that can reliably illuminate a teacher’s thought process.16 They were then given the study case vignette, describing a hypothetical attending rounds during which a frustrated intern speaks angrily about a patient. The intern refers to the patient, an injection drug user, first as a “shooter,” then as a “dirtball,” and suggests that the team move on to “someone interesting.” Attending physicians were asked to read each vignette fragment aloud, and to say all they could think of about what they would be thinking, saying, and doing were this situation actually taking place. Following this exercise, an open-ended interview was used to explore expressed attitudes and beliefs about clinical teaching and, in particular, about the role of the attending physician in teaching attitudes.
The principal investigator administered a semistructured interview 15, 17 to all 23 team members. Interviews included question about the roles of team members, clinical teaching, attitudes toward patients, and the teaching of attitudes. Participants were asked whether they had observed or undertaken efforts to teach attitudes, what seemed to work, and why. Incidents observed on rounds served as specific foci for interview questions. Interviews varied in duration from 40 to 130 minutes. Interviews, including think aloud exercises, were audiotaped and transcribed. Available hospital charts were reviewed for all patients cared for by study teams during the period of observation. We looked particularly for normative or judgmental statements about patients, explicit mention of patient comfort or dignity, and reference to behaviors of team members.
A conceptual model was developed based on principles of social interaction,18 and from it a list of themes was generated. Three investigators (JHB, DMI, JDC) used early interview transcripts and field notes to revise the theme list and create a preliminary coding scheme.19, 20 Coding was by thematic rather than linguistic units: an utterance or observation of any length could be coded as expressing or referring to a single theme, and a single complex utterance or observation might be coded as expressing several different themes.
Selected transcript fragments, with identifying information deleted, were then independently coded by each of these investigators. Differences were resolved by consensus, and the list of codes and coding rules were modified accordingly. Preliminary themes, patterns, and hypotheses were compared and contrasted both within and across cases, treating each ward team as a case. Feedback was elicited from clinical teachers, including research participants, regarding the plausibility of the emerging interpretations. The primary interview and observation data were then recoded according to the final coding scheme. Representative excerpts, edited slightly for readability, were selected to illustrate thematic findings.
While observations from all sources were combined to develop and test interpretations, findings were of two general types, corresponding roughly to the two chief modes of data collection. Participant observation provided the only direct access to actual behaviors, and therefore, unless otherwise specified, all illustrative examples of learner behaviors and attending physician responses were ones noted during participant observation. Conversely, we relied on data from interviews and think aloud exercises to gain access to participants’explanations and interpretations of their own and others’ behaviors. Our interpretations thus derive from taking these commentaries at face value, but also from considering them within the context in which we observed the behaviors to occur. Chart records were notably devoid of any reference to patient-centered attitudes or behaviors, and are therefore not cited.
Attending physicians described three categories of behaviors as “raising red flags” about possible deficiencies in concern, respect, or compassion for patients: (1) displaying reluctance to pursue clinically appropriate diagnostic and therapeutic steps, including avoiding admission, pressing for premature discharge, or otherwise cutting corners; (2) referring to patients in disparaging or derogatory ways, or otherwise showing disrespect; and (3) showing outright hostility, malice, or rudeness. Although the frequency of these behaviors is subject to interpretation, attending physicians agreed that cutting corners is quite common. Disrespect and hostility varied widely by team, with the tone set largely by the senior resident: on Dr. Dunn’s team virtually no episodes were observed, while on Dr. Baker’s, there were typically at least five or six such expressions per hour of rounds.
Attending physicians who noticed these problematic behaviors most often displayed no observable response. This finding was confirmed throughout our observations, by the attending physicians’ own accounts, and in interviews with residents and students. Attending physicians appeared more reluctant to intervene in practice than they described themselves to be in the think aloud exercise. Dr. Arthur, for example, took umbrage at the use of the term “shooter” in the think aloud exercise, and said he would stop rounds and “lay down the law” if ever an intern called a patient a “dirtball.” But he showed no discernible reaction when team members regularly referred to injection drug users as “shooters,” or when his resident described obese patients with pulmonary disease as “slug-like.”
Attending physicians often described themselves as responding silently, using nonverbal cues. Dr. Cooper described using body language to respond to derogatory comments:
I don’t say, “I don’t want to hear that stuff.” But I’m a pretty outgoing person, and when I don’t want to hear something, I become pretty rigid. It sends a message, a nonverbal but intentional message. It’s a physical demeanor. I’ll look directly at someone without the slightest hint of a smile and then change the subject. I think that tells them. I think they pick it up.
Such nonverbal responses were generally not apparent to the researcher.
A direct verbal response to problematic behavior was observed fewer than 10 times in 80 hours of attending-team interaction. When attending physicians did intervene verbally, they did so in one of three ways: by using humor, by casting feedback in terms of the learner’s self-interest, and by medicalizing interpersonal issues.
Using humor typically involved expressing dismay, but doing so jokingly, creating the impression that any breach of expected behavior had not been a serious one. For example, the senior resident on Dr. Baker’s team was informed during attending rounds that a patient who had left the hospital against medical advice, with an indwelling central venous catheter and serious pneumonia, had now returned to be admitted for treatment. The resident said: “I think we should get an ethics consult to find out if we even have to treat him, since he’s already broken a physician-patient contract.” Dr. Baker exclaimed, “That’s harsh!” and laughed. The rest of the team then laughed, and the conversation moved on.
In discussing this incident, Dr. Baker said:
I think he was saying it for effect. I don’t think he would do it. I think he was sort of testing the waters. But I think that our reaction was, “Well, of course we’d treat him.” I think that my response in saying “That’s harsh” sort of negates his statement, pretty effectively. So I wouldn’t think his statement would have very much of an effect, because it’s been canceled out.
The resident, when asked about this incident, interpreted Dr. Baker’s laughter to mean that she shared his negative feelings about admitting this patient.
The second observed verbal technique was to frame corrective feedback in terms of the learner’s own self-interest or self-protection. Typically speaking from personal experience, the attending physician suggested to the learner that behaving differently would ultimately save effort, prevent unpleasant conflict, or avoid embarrassing criticism. An example arose when Dr. Baker and her fourth-year student were waiting for the elevator, next to a woman in street clothes who appeared to be a visitor:
Dr. Baker: A pretty typical on-call night: pneumonia, endocarditis, TB … No cellulitis?
Student (loudly): We had one, but we got rid of him!
Dr. Baker (under her breath): Don’t say that, say ‘discharged.’ I got called on that once for saying it on the elevator, and dragged on the carpet by the attending.
Dr. Baker later commented:
I didn’t want her to be in the same situation that I had been in, of having someone point out, “You said the wrong thing in front of people who wouldn’t understand, and that was an error.” I didn’t think less of her for it. But I don’t want her to be in a situation of having to be called on the carpet for something like that.
Dr. Baker did not see herself as having pointed out that her student had said the wrong thing. The student perceived Dr. Baker’s cautionary tale as a friendly effort to save her from undeserved grief in the future. When asked what lesson she had learned, the student reported that she ought to be more careful in her word choice in public.
The third technique was to medicalize interpersonal issues. This involved framing feedback in terms of the ostensibly more objective, biomedical outcomes of care, and treating compassionate care as a means to those biomedical ends, rather than as an end in itself. For example, Dr. Cooper told her team that showing empathy for patients would increase honest disclosure and improve compliance, whereas hostility would produce poor information, and thus impede accurate diagnosis. Attending physicians worried that antagonistic attitudes might result in patients getting suboptimal technical care, or might be a marker for a higher probability of such care. Along these lines, Dr. Dunn saw the need to comment on attitude only if “really slovenly, poor interpersonal skills … interfered overtly with patient care.” It is noteworthy that he equated attitude with a set of interpersonal skills. Furthermore, he explained that “interfering with patient care” meant increasing the risk that a patient would not get the appropriate tests and treatments done.
More direct responses to problematic behaviors were not observed. Students and residents unanimously agreed that attitudes toward patients were never explicitly discussed on the wards. Dr. Arthur’s resident, when asked whether anyone talked on the wards about their attitudes toward patients or the way they treat patients, replied: “No. Nobody ever talks about that. Not at all.” Attending physicians never referred on rounds to moral principles, medical ethics, or professional norms. Although attending physicians commonly described hypothetical situations in which they would stop rounds and explicitly “lay down the law,” we witnessed nothing like this in 160 hours of observation. Although attending physicians occasionally offered nonspecific positive feedback, as when Dr. Baker told her interns that they seemed “good with patients,” we rarely saw praise or criticism of specific behaviors. And although attending physicians said they sometimes tried to model alternatives to problematic behaviors, they never explicitly told their learners what they were doing, or why.
Attending physicians had the opportunity to observe only a minuscule fraction of learners’ direct interactions with patients, typically in the relatively formal context of attending “walk rounds.” They were reluctant to pass judgment on attitudes based on the limited data they did have. There was similarly little opportunity for criticism that would not be disgracefully public. Some attending physicians attributed residents’ negative attitudes toward patients in part to their youth, naiveté, and relatively uniform social background. Finally, attending physicians sympathized with learners’ stress, were inclined to see negative attitudes as understandable expressions of it, and hesitated to judge those who were subject to inhuman working conditions when they, themselves, were not. Dr. Cooper commented:
I’m just not so angry at patients. But I also get at least 7 hours of sleep. And I’m not being called up at 4 am; I’m not being spit up on by another person; I’m not risking sticking myself with a needle. It’s a dangerous business, and they’re the ones on the front line.
Attending physicians found negative feedback of any kind difficult and unpleasant to give. They described having no training in doing so, not relishing the task, and often trying to avoid it. Most believed that residents’ personalities are largely fixed by the time they begin their internship. Dr. Arthur stated:
If somebody hasn’t gotten the message by the age of 25 or 30, they’re not going to get it now. I think you just try and model the behavior. But I don’t spend a lot of time dwelling on respect, or giving a ‘Sermon on the Mount’ about respect.
The reluctance to “preach” was buttressed by the belief that it would be inappropriate to impose on learners the attending physician’s own, possibly idiosyncratic, personal or political values.
The tendency to view learners as either good or bad further reduced the perceived efficacy of responding. Attending physicians believed that if a resident is “good,” and the observed behavior is only a lapse which does not accurately reflect underlying attitudes, then corrective feedback would only serve to discourage and frustrate the learner, perhaps even leading to more negative behaviors. Conversely, for residents who are fundamentally “bad,” attending physicians saw corrective feedback as futile, and instead emphasized damage control. Dr. Baker described her response to one such resident:
What I ended up doing was just removing him from the situation as much as possible, because there wasn’t any way I could see that I could change his behavior. He wanted to go to a conference, and I said, ‘Fine, why don’t you just go.’ I was just so glad to have him out of the hospital for a few days that I gladly did his work. I don’t think I gave him feedback. I just tried to hold my breath and get through the month.
Finally, even when they thought they ought to respond, attending physicians felt there was little in it for them. They were reluctant to alienate, offend, or incur the dislike of learners, and thereby risk receiving negative teaching evaluations. They also noted that besides simply being more pleasant, a harmonious working relationship was important to their peace of mind over the course of a month in which residents serve as their proxies. Furthermore, they did not feel supported or rewarded by their institutions for trying to teach attitudes. They believed that doing so not only would not advance their academic promotion or success, but might even open them to personal and legal risk. They saw the present cost-conscious academic medical environment as having little concern for teaching generally, and even less for teaching in noncognitive areas in which outcomes are difficult to measure. Dr. Arthur spoke of the increasing pressures to reduce length of stay, commenting:
Nobody is going to evaluate me on positive attitudes toward patient care. “Was respectful of patients”—I doubt that’s on a teaching evaluation. There will be all sorts of little questionnaires that people will develop for customer satisfaction, but this sort of modeling of behavior is not part of the economic question.
Our most striking finding is that attending physicians usually did not respond in any observable way to the problematic behaviors they identified as red flags for negative attitudes. A precise rate of responding cannot be meaningfully calculated, as both numerator (responses) and denominator (problematic behaviors) are subject to interpretation. But these attending physicians certainly let the great majority of these incidents pass without comment. When they did respond, they relied on nonverbal cues and on three indirect verbal techniques: humor, appealing to learners’ self-interest, and medicalization of interpersonal issues. The generalizability of the particular techniques we observed may be limited by the small number of teams studied in a single setting; however, their common features are consonant with the difficulty of giving specific, critical behavioral feedback.212,3,–24 Attending physicians saw explicit talk about attitudes as personal, punitive, and likely to be pointless, and as contrary to their own best interests. The feedback they did offer was subtle and often appeared to go unnoticed or be misinterpreted by learners.
The observed verbal feedback techniques avoid moral language, do not address or assume underlying negative attitudes, and leave room for face-saving positive reinterpretations of the precipitating problematic behavior. There are positive aspects to this strategy, but also hazards. Humor, for example, may convey the impression that the attending physician believes the behavior to have been meant jokingly in the first place, or is at least willing to give the benefit of the doubt to that interpretation. The door is thus opened to a tacit social reorientation whereby learners can change course and behave as if that indeed was what they had meant all along. Unfortunately, as we observed, learners may also interpret humor as endorsing their actions.
Appealing to self-interest cushions implicit criticism in an equally ingenious, though perhaps inadvertent, fashion. Far from being reprimanded for being insensitive to patients, the implication is that the learner, in her hard-working zeal, has neglected herself by behaving in a way that might bring criticism on her. Self-sacrifice and self-neglect in the pursuit of excellent patient care are powerful shared values in medical training.21 A potential moral criticism is thus given a subtle “positive spin” and transformed into moral praise. It is well established that negative feedback may be more acceptable and effective when accompanied by positive feedback.22 But does such a remarkable transformation of negative feedback into positive feedback represent a legitimate pedagogic tool, or simply avoidance of a difficult but essential task in professional education?
Medicalization serves similarly to put the learner’s motivations beyond reproach, and to move the content of feedback into a domain in which teacher and learner feel more comfortable. Human interaction is discussed, not as a goal in itself, but as a technique that novices must master in order to optimize data collection and patient cooperation. This orientation permits criticism to be recast in terms of the culturally endorsed aim of excellent technical patient care. Behaviors open to criticism are not the products of negative attitudes, but are simply errors. In Bosk’s terms, normative errors—unacceptable lapses in attitude, effort, or obedience to group norms—are reframed as nonnormative ones, as understandable and expected errors in technical performance, resulting from lack of skill and amenable to improvement with experience.25 Such skill-based criticism may be far less threatening to both teacher and learner. But if concerns about how patients are treated are not to be normative in medicine, what is? Translating feedback about attitudes and demeanor into blander, more palatable terms may leave learners with the impression that there are and can be no meaningful standards of attitude or value in the profession.
While attending physicians had well-intentioned and humane reasons for not directly confronting problematic behaviors, their confidence that nonverbal and indirect feedback techniques adequately conveyed their concerns appeared misplaced. Feedback is more readily accepted when it is low in inference—that is, aimed at observable behavior rather than inferred motives.21 Attending physicians may have operated intuitively according to a similar principle. But while their oblique techniques imply little or no judgment on the teacher’s part, they require extremely high degrees of inference by learners for their intent to be understood. Ende et al. found that preceptors in an ambulatory setting used similarly complicated and indirect feedback methods that minimized exposing learner errors.23 They point out that although indirect, nonconfrontational feedback helps preserve learners’ self-esteem, it may require too much inference on the part of learners, and therefore may not reliably provide them with the information necessary for accurate self-assessment.23 Ende et al. described this as “vanishing feedback”: the well-intentioned teacher, sensitive to the impact of criticism on the learner, offers feedback so indirect that nothing of value is transmitted.24
Attending physicians predicate much of their reluctance to offer direct feedback on a set of faulty assumptions. First, contrary to the belief that personalities are immutable, the years of early adulthood may be a time of tremendous moral development and attitudinal change.8 The very real pedagogic difficulties of engineering attitudinal change, and the moral difficulties of choosing the changes to encourage, should not be confused with an imagined psychological impossibility of change itself. Second, in the absence of any model of effective moral pedagogy, attending physicians turn to a passive conception of role modeling, which simply demands performing the desired behaviors oneself in the hope that learners will somehow absorb them. But compassion and respect are not discrete, specifiable behaviors; rather, they are expressed in highly complex and contextualized social interactions. Both theory 26 and empirical evidence 27 suggest that role modeling in such complex situations is more effective when teachers call attention to what they are modeling. Attending physicians may need to tell learners explicitly that they value the compassionate treatment of patients, and may need to narrate or review specific instances of deliberate modeling: “Notice that I always pull the curtain before examining the patient”; or “What I was trying to do in there was …”
Third, feedback can be direct yet remain nonjudgmental and positive. Feedback should be ideally descriptive rather than interpretive or evaluative, be undertaken in a collaborative spirit, and offer timely, brief, specific, performance-related information based on first-hand observation of remediable behaviors.24 Feedback, both positive and negative, should therefore be offered in response to individual episodes, not to overall patterns of behavior. Attending physicians could uncouple feedback about problematic behaviors from judgements about good or bad character, either by focusing on observable characteristics of the behavior, or by describing their own emotional responses.22 For example, rather than say, “That was disrespectful of you,” one might say, “You used a term which many people find offensive,” or “I really don’t like that word, because …”
Several institutional changes might facilitate more effective responses by attending physicians. Clinical teaching can be restructured to provide more time to observe learners interacting with patients and more opportunity for longitudinal teacher-learner relationships that might provide a safer context for meaningful feedback. Second, faculty development programs can provide opportunities to learn and practice concrete skills such as giving specific feedback. Third, the dimensions of attitude and role modeling can be built explicitly and seriously into the evaluation and promotion of both faculty and trainees. Fourth, making professional attitudes a frequent and legitimate topic of explicit discussion may help demonstrate their importance. Feedback about attitudes takes place in the context of the “hidden curriculum” that expresses the values of the institution.28 Commitment to compassion and respect must therefore be given conspicuous and credible priority by those in positions of influence.
Finally, perhaps the most important type of modeling is to treat learners with the compassion and respect with which we want them to treat their patients.29 Although there is room in such relationships for discretion and reserve, a high premium must be placed on forthrightness. Respectful professional education provides learners the opportunity to see clearly when their behaviors do and do not meet professional and ethical expectations, and assists them in shaping an identity of which they can be proud.
The central dilemma illuminated by our findings is whether the indirect responses observed are adequate to shape learners’ professional attitudes, or whether more explicit feedback and talk about values are called for. Our observations suggest that attending physicians are missing priceless teaching opportunities, albeit with commendable intentions. We propose that the values to which medicine professes should be made explicit to learners, not swept under the rug. In the absence of outcomes data regarding changes in learner attitudes and behavior, we acknowledge that there is room for honest disagreement over the best ways to provide moral guidance. We hope that our findings spark explicit debate among teaching faculty, and between faculty and learners, regarding the desirability of various attitudes and behaviors, the appropriateness and effectiveness of different kinds of feedback, the nature and impact of modeling, and the content of the hidden curriculum. What is not debatable is our responsibility as teachers of medicine to take seriously the transmission of professional values.
Dr. Burack was supported by a National Research Service Award, U.S. Public Health Service (5 T32 PE1002). This research was further supported by cooperative agreements (1U76MB 00005-01 and -02) with the Bureau of Health Professions, Health Resources and Service Administration, U.S. Public Health Service.
The authors thank Professors Sam Wineburg and Patricia Phelan and the students of Educational Psychology 586 and 587 at the University of Washington in 1995, and the members of the 1994–1996 University of Washington National Research Service Award Works in Progress Seminar for their helpful advice and input, and especially the study participants for their forbearance and openness. We also thank several anonymous reviewers for their thoughtful and constructive suggestions.