McKeown, Robert E. PhD; Reininger, Belinda M. DrPH; Martin, Maurice PhD; Hoppmann, Richard A. MD
Over the past three to four decades, there has been significant interest in the patient—physician relationship, especially as it relates to decision making.1,2 The period is characterized by a gradual shift away from a paternalistic model, in which beneficence is the dominant operative principle, toward models that give greater weight to autonomy and self-determination by the patient and that see beneficence as one of several principles guiding decision making. There is, thus, greater interest in shared decision making, i.e., decisions made jointly by the patient and the physician. Much of the current research regarding the physician—patient relationship examines the respective roles of the patient and the physician in the areas of understanding risks versus benefits of treatment and procedures and who controls the decision-making process. These investigations show that both physicians and patients assume varying levels of responsibility and control of the patient's health care.3–6
One theoretical model of the physician—patient relationship7 describes four distinct styles of relating that have implications for communication and shared decision-making and treatment decisions.
* Paternalistic styles have the physician in the role of expert, with decisions about the patient's health care controlled by the doctor. The attitude of the patient is: “Whatever you think, you're the doctor.”8
* Mutualistic styles employ the patient and physician as equal partners in the decision-making process; e.g., the physician may say: “I want you to know the options so together we can make the best decision for you.”9
* Consumeristic styles put the patient in a role with greater control of decisions as if he or she were purchasing a health care product. They may say: “I want all of the information about my problem and will consider a second opinion before I decide what I would like to do.”10
* Default styles are characterized by neither the patient's nor the physician's taking responsibility for the decisions about the patient's health care. Instead, the course of action is dictated by a third party or determined by extraneous factors: “Mr Jones, your insurance company will allow only these tests to be done at this time.”11
Among other similar models of patient—physician decision making are the “deliberative” and “interpretive” approaches described by Emanuel and Emanuel.2 In the deliberative approach, a variant of the mutualistic style, the physician advocates for a particular decision based on knowledge of the patient's values and condition and the options. The physician might say, “The final decision is yours, but knowing your values and concerns, I believe this is the best option for you and here's why.” In the interpretive style, a variant of the consumeristic model, the physician plays a more advisory role, but still leaves the final decision in the hands of the patient. Thus, the physician may say “Let's clarify what is most important to you and I will help you to know and understand how the options relate to that so you can make the best decision for yourself.”
Some research of patients' decision making concluded that patients rarely relinquished their role entirely.6 An examination of the decision to perform amniocentesis concluded that the patient's preference, not the physician's, was the most significant factor finalizing the decision.12 Other studies have emphasized the complex, multifactorial, and variable nature of the decision-making process in health care, as in the case of gynecologists' attitudes toward hysterectomy decisions.13
The present study, a part of an examination of ways to evaluate and enhance the graduate medical curriculum, examined the views of first-year residents and residents' clinic patients concerning the role each expected to exercise in making health care decisions, as well as the role each perceived the other to want in the process. A greater understanding of these perceptions of the patient—physician relationship has implications for educating physicians and, ultimately, for the quality of the patient—physician relationship.
Study Design and Population
The data-collection procedures were reviewed by the University of South Carolina Institutional Review Board and the Hospital Review Committee. All data were collected by graduate students who were trained in survey protocol and procedures.
All new residents entering training at the University of South Carolina School of Medicine and Palmetto Health Alliance Program were surveyed during their orientation session in the summer of 1999. The present analysis is limited to first-year residents in seven programs: internal medicine, family medicine, pediatrics, obstetrics and gynecology, psychiatry, emergency medicine, and surgery. Programs that had only one first-year resident and those less likely to encounter the types of scenarios used (dentistry, orthopedics, ophthalmology) were excluded. Of the 72 residents' surveyed, the 45 responses used here were from first-year residents in the seven programs, with information about race and gender and most scales completed.
Patients were interviewed during the fall of 1999 at four university clinics: obstetrics and gynecology, pediatrics, internal medicine, and family medicine. The pediatrics clinic interviews were not included because the scenarios presented to parents or guardians in the pediatrics clinic differed from those presented to adult patients. (Pediatrics residents were given the instrument corresponding to that for adult patients.) Office receptionists identified all patients who came to see residents for treatment, and the patients were approached in the waiting room and asked to participate in the study. If the patient consented after being further informed about the study, a 25-minute structured interview was conducted. The structured interview was used to improve data quality in this patient population, which includes persons of low literacy levels. We believe the self-administration by residents and structured interview of patients improved the quality of the information obtained from both.
Patients were not screened for disease states and were not asked about their own health. A total of 67% of patients approached at the clinics consented to participate. Among the stated reasons for not participating were time constraints, physical discomfort or inability, and language barriers. Some did not disclose reasons for refusal. Of the 123 questionnaires collected from patients, 92 were from the three clinics and provided enough information for at least partial analysis.
To construct the instruments (the residents' questionnaire and the patients' structured interview), we used a modification of the eight-item Information-seeking (IS) scale and the six-item Decision-making (DM) scale from the Autonomy Preference Index (API).14 The original instrument had three scenarios; we retained only one concerning hypertension and added four scenarios concerning depression, cholecystectomy, hysterectomy, and possible prostate cancer. All IS and DM items used a Likert scale, coded so that stronger preferences for patients' control of decision making or information seeking received higher scores. As in the original API, some items were reversed for scoring, and the scores were converted to a 0-to-100 scale.
The residents' paper-and-pencil questionnaire took approximately 30 minutes to complete and contained six questions about demographics and 113 questions dealing with decision making. The patients' structured interview questionnaire contained 76 items for adult men patients and 79 items for adult women patients, including four items for demographics and scales concerning decision making. Patients were not asked to respond on the basis of their experience with a particular physician, but were asked, “What do you think ‘your doctor’ would say?” The hysterectomy scenario was asked only of women patients and the possible prostate cancer scenario was asked only of men patients. Because obstetrics and gynecology patients are exclusively women, the internal medicine and family medicine clinics were oversampled for men.
The API was further modified to be administered to patients and resident physicians. For both groups, respondents were asked to indicate their own views, and then to indicate what they believed the other person in the patient—physician relationship preferred. Examples of questions include: “When patients are in the hospital, they should not be making decisions about their own medical care” and “A physician should explain the purpose of his/her patient's laboratory tests.” For the scenarios, higher scores indicate the amount of control desired by or attributed to the indicated person. In four vignettes, both respondents were asked what control they believed a consultant (such as a surgeon) would want. A score of 50 represents equal participation or control of the decision by both physician and patient, as in the mutualistic model. A score of 0 indicates the decision is made completely by others, 25 indicates the decision is made mostly by others, 75 indicates the decision is controlled mostly by the person indicated, and a score of 100 means that person alone makes the decision (a paternalistic model if that person is the physician, a consumeristic model is that person is the patient). Differences of 25 or more indicate major shifts along the continuum of decision-making control, while differences of ten to 15 points indicate a less substantial, but still meaningful, difference in the balance of control.
Statistical analysis examined differences in scores between patients and residents and differences in scores from the same respondent (within-subject differences), that is, the difference in the self-reported score and the score attributed to the other person. To compare scores for patients with scores for residents, we employed analysis of variance (ANOVA), controlling for race and gender. We used a repeated-measures ANOVA, controlling for race and gender, to test for statistically significant within-subject differences. Similar procedures were used to examine scores attributed by both sets of respondents to the consultants. Unadjusted mean scores and standard deviations were presented.
Table 1 presents basic demographic characteristics of the residents and patients: 50% of the residents were women, 78% were white, and 87% were between 18 and 30 years of age; 62% of the patients were women, 61% were African-American, and the age distribution was fairly even across adult age groups. Residents in the three clinics from which patients were drawn constituted almost half (46.7%) of the residents in the analysis.
API Basic Decision Making and Information Seeking
Decision making. Table 2 compares patients' and residents' general perceptions about roles in decision making and information seeking both for themselves and what they perceived the other would want. None of these questions was tied to any specific medical scenario or condition. The patients wanted slightly less control than the residents thought they did (41.4 versus 53.0, p = .0130), and the residents thought patients should have substantially greater control than the patients assumed physicians would want for them (64.6 versus 32.9, p < .0001). Although the residents thought patients wanted somewhat less control than they were willing to give them, the difference was not statistically significant at common alpha levels after controlling for race and gender (53.0 versus 64.6, p = .1452). Conversely, the difference in control patients wanted for themselves compared with what they believed their physicians would want for them, while not great, was statistically significant (41.4 versus 32.9, p = .0008).
Information seeking. Table 2 also shows that both residents and patients felt strongly about the need to share information with the patient (i.e., all high scores). Although there are statistically significant differences between residents and patients and within patients, these differences do not indicate clinically meaningful differences in perceptions of the proper or desired amount of information to be shared. The greatest difference is less than six points (although that difference has the smallest p value). The critical difference, between a patient's desired level of information and a resident's desired level of information sharing, is less than one point in mean scores.
Unlike the DM and IS scales, where all scores represent the degree of patients' control, the self-reported scores in the scenarios represent the amount of control the respondent desired, and attributed scores represent the amount of control the respondent believes the other person would want. See Table 3.
Hypertension. There was virtually no difference in the scores for the desired amount of control patients expressed and the amount attributed to them by the residents (52.4 versus 56.1, p = .8846). Similarly, there was little difference in the degree of control residents desired and the amount patients attributed to them (75.3 versus 68.3, p = .1595). Both patients and residents gave higher control scores to physicians than to patients, substantially more in the patients' responses (p = .0008), and about half as much difference among residents (p = .1455).
Depression. There was essentially no difference between patients' and residents' perspectives on either patients' or physicians' desired control of decisions concerning depression (p = .1892 and p = .9228). The intra-subject differences, although greater, were statistically significant only among residents (p = .0066). In this scenario, both patients and residents attributed a greater desire to control decisions to the physician. The degree of control attributed to the consultant by the resident was substantially and significantly greater (p < .0001), while the patient saw no difference in desired control between the consultant and the physician (p = .9811).
Cholecystectomy. In this scenario patients believed physicians wanted much greater control of the decisions than the residents actually asserted for themselves (73.0 versus 41.3, p < .001). Conversely, residents believed patients wanted somewhat less control than the patients asserted for themselves (57.8 versus 68.6, p = .0059). Although the control patients said they wanted and what they thought their doctor would want were not significantly different, residents assumed patients wanted more control than the residents asserted for themselves (57.8 versus 41.3, p = .0035). The residents again assumed the consultant would want considerably more control of the decisions than patients or residents (p < .0001), while the patients again saw no difference in the desire for control of their physician or the consultant.
Hysterectomy. In this scenario, women wanted greater control of decisions than in all the other scenarios except cholecystectomy, but they also attributed a desire for the same amount of control to the physician and the consultant (mean scores 71.3, 74.4, and 73.6, respectively). Residents wanted substantially less control than they attributed to their patients (37.2 and 57.7, p = .0008) and far less than attributed to them by patients (37.2 versus 74.4, p < .0001). The residents again assumed that a consultant would want much greater control of decisions than would either the patient or the residents themselves (76.3 versus 57.7 and 37.2, p < .0001 for resident compared with the consultant).
Prostate cancer. Both patients and residents assumed significantly greater control preference for physicians compared with patients (76.1 versus 51.4, p = .0181, and 64.6 versus 46.2, p = .0141). Patients' and residents' scores for the patient's desire for control were not significantly different (51.4 versus 46.2, p = .9498). Residents wanted slightly more control for themselves, though less control than what patients attributed to them (64.6 versus 76.1, a difference that may be clinically important, though not statistically significant, p = .0701). The residents again attributed greater desire for control to a consultant than they did to themselves (77.2 versus 64.6, p < .0001), while the patients again attributed equal amounts of desire for control of decisions to the physician and the consultant.
The control of decision making and the sharing of information in medical situations are complex topics, but they are fundamental to the patient—physician relationship. Understanding the dynamics of the relationship, including the role each partner plays and his or her perceptions of the other partner, has implications for improving the quality of the patient—physician relationship and improving the medical education process.
In our study, residents were willing to give far greater control to patients than the patients expected them to give, and the residents assumed patients would want a slightly greater level of control than the patients actually claimed for themselves in nonspecific situations. In the scenarios, the scores changed and some patterns shifted, indicating that responses differed when the more concrete scenarios replaced the more general-scale questions that lacked a context. Both residents and patients agreed on the amounts of information sharing they wanted and thought the other would want. While there was mutual understanding and agreement about information sharing, there were real differences in what residents and patients perceived as their decision-making roles in the physician—patient relationship when the questions did not relate to a specific medical situation or condition.
The patients' higher scores for information-seeking topics as opposed to decision-making topics reflect a strong desire by the patients “to know,” which is consistent with results found by Deber and colleagues using a similar instrument.6 Also, as in the Deber et al. study, we found that the relative roles of patients and physicians varied by the scenarios presented. Three patterns emerged from the case scenarios. For the hypertension and depression scenarios, patients and residents attributed greater control to the physician, but there was no significant difference between the degree of control they wanted for themselves and the degree of control the other attributed to them. However, the patients' desire for equal sharing in the decision making may be difficult to reconcile with the residents' desire for more control. Thus, there appears to be a fairly clear understanding of what each expected in these two scenarios, although the expectations were not entirely compatible. From the patients' perspective, these responses were examples of mutualistic9 decisions, whereas the residents' perspectives were more akin the deliberative model,2 or perhaps even a traditional paternalism.
In the cholecystectomy and hysterectomy scenarios, the patients did not attribute a significant difference in the degrees of control for themselves and the residents, but the residents did. Patients wanted somewhat greater control of decisions for themselves, but wrongly assumed their physicians wanted the same or perhaps more control. In fact, in contrast to the hypertension and depression scenarios, the residents' responses to the cholecystectomy and hysterectomy scenarios were to give greater control to the patient (residents' scores <50). In this scenario, even though each group's perception of the other was mistaken, the residents' desire for less control and the patients' corresponding desire for more control were compatible and appeared to fit the interpretive model2 well.
In the possible prostate cancer scenario, both patients and residents perceived the physician to be playing a greater role in the decision-making process, while the patients and residents both saw the patient as sharing (or wanting to share) equally in the decision. Unlike the responses to the cholecystectomy and hysterectomy scenarios, however, the residents did not assume patients wanted a greater role in the decision making. In addition, there was no significant difference between the role the patients wanted to play and the role the residents attributed to patients. Residents accurately assessed the patients' desire for equal control while also believing a greater amount of control for themselves was appropriate. This scenario corresponds more closely to that of the mutualistic model from the patients' standpoint, but slightly favors the deliberative model from the residents' perspective.
All ANOVA models used to test within-subject differences and differences between patients and residents controlled for race and gender. In all but a few of the models, there was no indication of significant differences in scores by race or gender. The age range was too narrow among residents to be meaningfully used as a covariate, and adding age group to patient models did not alter the results. To make them consistent with the residents' models, the patients' models are reported without controlling for age. Because our focus was on comparisons of residents and patients and we controlled for race and gender in evaluating those comparisons, we did not report the few instances in which race or gender was statistically significant.
Our study's findings suggest that patients and residents assume a number of different roles with respect to decision making, and that these roles depend on the health care scenario. For example, one possible explanation of the relational patterns we observed is the perceptions of the acuity and severity of the medical situation by those involved. Compared with the cholecystectomy and hysterectomy surgeries, hypertension and depression could be considered to be more chronic medical problems, with effective treatments requiring long-term adherence by patients involving unpleasant side effects. By contrast, in the more acute scenarios, the patient might expect to have the same level of decision making as the physician, and the physician might want the patient to assume a relatively greater role because of the seriousness of the situation and the potential for adverse outcomes inherent in surgery, especially for hyster-ectomy, often an “elective” procedure.
The possible prostate cancer scenario actually falls somewhere in between the medical and surgical scenarios, which may be the reason patients attributed greater control to the physicians than to themselves. Because only men patients responded to this scenario, the results could also reflect differing attitudes of men and women in medical situations pertaining to themselves. However, on the basis of post-hoc analysis, the fact that women outnumbered men in the patient sample does not entirely account for the concordance between the hysterectomy and cholecystectomy surgery scenarios (the former answered only by women) and the difference from the possible prostate cancer vignette (answered only by men; data not shown). In addition, we did not ascertain whether any of the patients had had personal experiences with the medical conditions described in the scenarios, so it is not possible to comment on what role personal experiences might have played in their responses.
The role of the consultant as perceived in the scenarios by the residents and patients was consistent and somewhat surprising. Residents assumed the consultant would want greater control than did the residents themselves, while the patients saw the physician and the consultant as playing equal roles in the decision-making process. This is a very important concept for physicians and patients to appreciate. It implies that when a consultant becomes involved in the case, the resident anticipates the consultant's playing the lead role in the decision-making process, whereas the patient sees the consultant as just another member of the physician team. These different views of the consultant could easily lead to confusion and miscommunication in the patient—physician relationship, especially as it relates to treatment.
One important contribution of this study stems from the characteristics of the patient population. More than half of these patients were African American, and a large proportion of patients served by these clinics are from low-income families. These results indicate that one should not assume patients from disadvantaged populations do not want to play a substantial role in making decisions concerning their health care. Physicians would do well to probe for the levels of control desired by patients from similar populations.
It should be noted that the residents' results are the attitudes of recently graduated medical students, not those of experienced residents or practicing physicians. Therefore, they reflect the influence of medical school education, especially the clinical years (three and four), and can be considered as representing the starting point of residency training. Future studies to address this and other limitations of this report should include surveying residents further into their training to assess the influence of greater experience on the patient—physician relationship; surveying residents from other training programs; matching residents to specific clinic populations as a function of their chosen areas of residency; and, after educational intervention, a follow-up survey of how the residents' understanding of and approach to the patient—physician relationship may have changed.
It would also be important to expand the patient population to “private practice” patients, who are generally better educated and at higher socioeconomic levels, to determine whether these factors are associated with a desire for information and involvement in decision making that is different from that of the patients in our sample. Also, gathering data from patients about previous experiences with men versus women physicians would be important in future research.
The physician—patient relationship is complex, but worthy of serious investigation in the hope of improving our understanding and better preparing our young physicians to engage in positive and meaningful communication with their patients. Furthermore, sharing the findings of such studies with patients, health care administrators, and others involved in the delivery and management of health care could likewise positively affect the physician—patient relationship and alter the environment in which that relationship takes place.