To achieve maximum treatment effectiveness, physicians must persuade patients with chronic diseases to commit to and follow through on treatment recommendations. Physicians who routinely involve patients with chronic diseases in treatment decisions (presenting options, discussing the pros and cons of those options, eliciting patient preferences, and reaching mutually agreed-on treatment plans) can be said to have a "shared" or "participatory" decision-making style. Such physicians may have greater success in securing patient cooperation—and therefore may have better patient health outcomes—than physicians with more controlling decision-making styles.

Although research has emphasized greater patient involvement in treatment decisions, less attention has been paid to the physician as an agent for obtaining this result. In studies of patients with hypertension [1], non-insulin-dependent diabetes mellitus [2, 5], peptic ulcer disease [3], and rheumatoid arthritis [7], patients whose physicians were less controlling (or more participatory) during office visits had better functional status and lower follow-up glycosylated hemoglobin levels, blood pressure, and arthritis severity than patients of less participatory physicians. These studies suggest that physicians vary widely in how much they facilitate patients' active participation in treatment decisions (what we have termed a "participatory decision-making style").

Our study was part of the Medical Outcomes Study, an exploration of differences in the organization of health care delivery and physician reimbursement mechanisms and specialty, and their effect on patients' health outcomes. We examined personal and practice characteristics of physicians likely to influence variations in participatory decision-making style, defined as the propensity to offer patients choices among treatment options and to give them a sense of control and responsibility for care. We relate physicians' background and training, sociodemographic characteristics, practice volume, and satisfaction with professional autonomy to the patient-reported measure of physicians' participatory decision-making styles, after adjusting for patient and practice characteristics.

Methods

Top Methods Results Discussion Author & Article Info References

Study Design

The study design and sampling strategies of the Medical Outcomes Study have been described [8, 9]. Data for this study derive from a cross-sectional sample of patients obtained from the offices of physicians participating in the Medical Outcomes Study during a 9-day study enrollment period in 1986.

Physician and Patient Samples

Physicians 31 to 55 years of age who were board eligible or board certified in general internal medicine, family practice, cardiology, or endocrinology were included. The resulting sample of 337 physicians was restricted to the 300 who completed a clinician background questionnaire. These physicians comprised 28 cardiologists (9.3%), 20 endocrinologists (6.7%), 79 family physicians (26.3%), and 173 general internists (57.7%). They were primarily young (mean age ±SD, 39.2 ± 6.3 years), white (83.7%) men (78.5%). One hundred fifty-five (51.7%) physicians practiced in solo or single-specialty group practices; 52 (17.3%), in multispecialty groups; and 93 (31.0%), in health maintenance organizations.

Of 22 463 adult patients of physicians participating in the Medical Outcomes Study, a random half were asked about participatory decision-making style on their screening questionnaire [10]. The patient sample for this analysis was further restricted to only patients (n = 7730) of the 300 physicians who completed the clinician background questionnaire.

The average age of patients included in this study was 46.5 ± 17.7 years (range, 18 to 106 years); 20.2% were 65 years of age or older. The patients had completed 13.7 ± 3.0 years of education; most (61.3%) were women, and 21.7% were nonwhite. More than half (53%) had one or more major chronic conditions [8].

Data Collection

Data for this study were taken from the Medical Outcomes Study patient and physician screening visit questionnaires and the clinician background questionnaire [10]. The three questions measuring participatory decision-making style were included on the patient screening visit questionnaire. Patients were asked to rate their physicians' style on a five-point scale, using the following questions: 1) "If there were a choice between treatments, would this doctor ask you to help make the decision? [definitely yes to definitely no]"; 2) "How often does this doctor make an effort to give you some control over your treatment? [very often to never]?"; and 3) "How often does this doctor ask you to take some of the responsibility for your treatment? (very often to not at all)?" We created a scaled score by calculating a sum of these items. We then transformed this scale to range from 0 to 100 by subtracting the lowest possible value of the scale from the raw score, dividing the result by the maximum possible range of scores, and multiplying that result by 100. The mean ±SD, averaged across each physician's patients, for this transformed three-item scale along with its internal consistency reliability are presented in Table 1. The average interclass correlation among scores for each physician was relatively high (r = 0.62; P < 0.001). The average number of patients providing participatory decision-making style scores for each physician was 25.8 ± 11.2.

Questions on the personal and practice characteristics of physicians, from the clinician background questionnaire, included practice volume (measured as the number of outpatient visits that physicians reported having in a typical week of their main professional practice), participation in primary care and interviewing skills training during residency, and satisfaction with professional autonomy. Items were rated on a five-point scale, ranging from very satisfied to very dissatisfied. On the physician screening visit questionnaire, physicians were also asked how long they spent face to face with each study patient for the office visit. Because physicians might see the same volume of patients distributed over more or less concentrated blocks of time, we also measured time spent per patient to assess practice volume. Analyses using time spent per patient compared with practice volume produced highly similar results [11].

Analytic Strategy

The analysis involved two stages. To minimize the effects of patient characteristics on physician style, we did a patient-level multiple regression analysis using patient-reported participatory decision-making style as the dependent variable and the patient's age, the patient's education level, the patient's sex, the patient's minority status, type of payment for health care, presence of eight selected common chronic conditions, the patient's perceived health status, and each physician's identification number (entered as dummy variables) as independent variables and covariates. We used individual physician identification numbers to remove the variation in physicians' participatory decision-making style that might have been due to unobserved physician characteristics. The results of this patient-level regression analysis [11] showed that patients who were older than 75 years of age or younger than 30 years of age, had less than a high school education, were not white, were male, had more comorbid conditions and poorer perceived health status, and reported a shorter relationship with the Medical Outcomes Study physician reported less participatory visits. Twenty-one percent of patients were seeing the Medical Outcomes Study physician for the first time, and another 6% were seeing the physician on an ongoing consultative basis. Because first and consultative visits were similarly distributed across the physician specialty groups and did not account for the observed differences in participatory decision-making style among the specialties, we retained these visits in the analysis.

To account for differences in these patient characteristics in Medical Outcomes Study physicians' patient samples, we used this patient-level regression model to calculate an expected participatory decision-making style score for each Medical Outcomes Study physician. We then subtracted this expected score from the observed score (patients' actual style reports) for each physician. In all analyses, we have used the difference between this observed and the regression-generated expected participatory decision-making style score, transformed to range from 0 to 100 using the technique described above, as the dependent variable in statistical models. The use of the observed score minus the expected score results in a narrow numeric range but minimizes the variation in participatory decision-making style that would otherwise have been due to differences in Medical Outcomes Study physicians' patient panels. Observed differences in participatory decision-making style related to physician characteristics ranged from 33.2% to 80.6% of the adjusted participatory decision-making observed style score minus the expected style score's standard deviation.

In the second stage of the analysis, we developed physician-level models, relating the observed score minus the regression-generated expected score to physician's age, sex, minority status, specialty, site of care, and type of practice (health maintenance organization, fee-for-service), entered as dummy variables, and statistical interactions between site and type of practice, to adjust for differences in practice and physician characteristics that varied by site and features of specific practice organizations (such as practice volume). The regression model also included practice volume, primary care or interviewing skills training (entered as a dummy variable), and satisfaction with autonomy. Data for Table 2 and Table 3 derive from this physician-level multiple regression model. Data for Table 4 were obtained by adding a specialty "by volume" interaction term to the multiple regression model.

We tested practice volume and average length of office visits in comparing participatory decision-making style by specialty. Cardiologists saw fewer outpatients per week and reported spending the most time per office visit; physicians practicing family medicine reported seeing the most outpatients per week and spent the least time per office visit. Using either practice volume or time spent per patient did not change the relation between specialties and participatory decision-making style. We thus used practice volume as a covariate in all analyses.

To facilitate presentation of the data, we re-ran the models using dichotomous versions of practice volume and satisfaction with autonomy. Although variability that resulted from using dichotomous variables was reduced, study findings did not change with varying specification of independent variables and covariates. We weighted each physician's style score by the number of patients sampled from the practice (mean, 25.8 ± 11.2), by the probability of the specialty being sampled within a practice organization, and by the proportion of the specialty in the U.S. physician population, using a weighting function described elsewhere [8]. Although this weighting did not change the substantive results, we report weighted results to maximize the generalizability of study findings to a broader physician sample.

Results

Top Methods Results Discussion Author & Article Info References

Two separate analyses based on data from the Medical Outcomes Study supported the validity of the patient-reported participatory decision-making style scale as an attribute of the physician. First, using a six-item self-reported measure of participatory decision-making style, included on the clinician background questionnaire, we found a statistically significant correlation between physicians' and patients' reports of participatory decision-making style (r = 0.45; P < 0.001, correlation attenuated for reliability of physician and patient participatory decision-making style estimates). Second, in a patient-level analysis, we correlated the patient-reported participatory decision-making style measure with the patient's satisfaction with the following: the physician's personal manner, the physician's explanation of what was done, the physician's technical skills, and the office visit overall (each item was rated on a five-point scale and was included on the patient screening visit questionnaire). The correlation between participatory decision-making style and the four-item patient satisfaction measure was statistically significant (r = 0.46; P < 0.001). However, participatory decision-making style was positively related to many independent variables (such as patient education and preference for involvement in care) that were negatively related to patient satisfaction, indicating that participatory decision-making style and patient satisfaction are conceptually distinct.

Distribution of Participatory Decision-Making Style Scores and Interpretation of Unit Differences

The observed score minus the expected score of participatory decision-making style, transformed to range from 0 to 100, had an actual range of 54 to 100 (mean, 80.06 ± 6.58). Of the 300 study physicians, scores for the highest quartile of participatory decision-making style scores (those rated as most participatory) were 84.2 or higher, whereas the scores for those rated in the lowest quartile (the least participatory) were 76.0 or lower.

Higher participatory decision-making style scores were associated with less changing of physicians by patients during a 12-month follow-up period. Roughly one third of the patients of physicians scoring in the lowest quartile of the participatory decision-making scale and 15% of patients who rated their physicians in the highest quartile changed physicians in this 12-month period. In a patient-level logistic regression model predicting change of physician in a 12-month period (with adjustment for patient age, education, sex, and minority and health status) participatory decision-making style was a statistically significant predictor of change (odds ratio, 0.71; CI, 0.68 to 0.89; P < 0.01).

Because using an observed minus expected score decreases the amount of variation available for explanation and complicates an otherwise intuitive interpretation of scores ranging from 0 to 100, small numeric difference in these adjusted style scores—of the magnitude of those presented here—were meaningful with respect to patient care. Differences of 2.00 scale points in the mean of the observed score minus the predicted score on the participatory decision-making style scale—a score of 81 compared with a score of 79, for example—equaled 31% of the standard deviation for this scale. Differences of 2 points were related to a 10-percentage point difference in the likelihood that patients would leave a physician's practice in the next 12 months. Small unit differences in other measures of interpersonal care, such as patient satisfaction with care, have been shown in other studies [12] to have important behavioral consequences, for example, a 10-fold increase in the probability of disenrollment from health plans.

Differences in Participatory Decision-Making Styles

Practice Volume

A more participatory decision-making style could require more time with the patient and could thus be constrained in a high-volume practice. We noted that physicians seeing more outpatients per week were rated as less participatory than those seeing fewer outpatients (r equals – 0.22; P < 0.001). Similarly, more time spent per patient was associated with a higher participatory decision-making style rating (r = 0.23; P < 0.01). These variables were highly intercorrelated, and total outpatient volume provided a broader estimate of total practice intensity; thus, we used outpatient volume in the following analyses.

We used the continuous version of the practice volume variable in the multiple regression model but, to facilitate presentation of study results, we dichotomized physicians' reports of the numbers of outpatients seen in a typical week at the median. As shown in Table 2, those in high-volume practices (more than equal to 70 outpatients per week) had lower participatory decision-making style ratings than those in lower-volume practices (P < 0.005). The magnitude of this difference equaled 44.8% of the standard deviation of the adjusted participatory decision-making style score. Similar results were obtained when the average length of the office visits reported by study physicians was used as the indicator of practice volume.

Physicians in each of the specialty groups represented in our sample reported substantial and statistically significant differences in outpatient practice volume, in both the number of outpatients they reported seeing in an average week and the average number of minutes they reported spending with each patient.

Family practice physicians reported seeing 90.4 ± 36.7 outpatients per week. This figure was 29% higher than that reported by general internists (70.1 ± 25.9 outpatients per week; t equals 6.02; P < 0.01), 35% higher than that reported by endocrinologists (66.9 ± 29.7; t equals 4.26; P < 0.01), and 115% higher than that reported by cardiologists (42.1 ± 26.9; t equals 9.25; P < 0.01). Family practice physicians reported spending significantly less time per patient, on average, than any other physician group.

Specialties

Cardiologists were rated as the least participatory physicians (observed style score minus expected style score, 78.18 for cardiologists compared with 80.69 for all other physicians; P < 0.05). When we included consultation visits (which were rated as slightly less participatory for all physician groups) in our analysis, cardiologists were rated as significantly more participatory on these visits than physicians in other groups, although including these consultative visits did not alter the relation between specialty and participatory decision-making style.

General internists, family practice physicians, and endocrinologists were rated as more participatory in their decision-making styles when in low-volume practices than in high-volume practices (divided at the median of 70 outpatients per week for ease of presentation) (Table 4). Differences for general internists and family practice physicians in low-volume practices compared with high-volume practices were 45.4% and 80.6% of the adjusted participatory decision-making style score, respectively. No statistically significant differences were seen between participatory decision-making style ratings for cardiologists in high- and low-volume practices.

Physician Characteristics

Physicians who reported training in a primary care-track program or who had training in interviewing skills during residency (n = 90) were reported by their patients to have more participatory decision-making styles than those who did not (n = 210; P < 0.05) (Table 3). The observed participatory decision-making style score difference between these groups was 33.2% of the adjusted participatory decision-making style standard deviation. The average number of years since primary care or interviewing skills training was 10 years. We tested for the significance of interactions between physician age and primary care or interviewing skills training and found that they were not significantly related to participatory decision-making style ratings.

Overall, physicians who reported that they were satisfied with the level of personal autonomy in their professional lives were rated as more participatory than those who reported that they were not (P < 0.05) (Table 3). Observed differences were 34.2% of the adjusted participatory decision-making style standard deviation. Participatory decision-making style increased with higher satisfaction with professional autonomy (r = 0.25; P < 0.05). Satisfaction with level of autonomy also decreased with increasing practice volume (average number of outpatients seen per week) (r equals – 0.19; P < 0.01).

Patients reported that nonwhite physicians had a less participatory decision-making style than white physicians (Table 3). Observed differences were 34.6% of the adjusted participatory decision-making style standard deviation. Nonwhite physicians also reported seeing more outpatients per week than did white physicians (82.1 ± 45.1 compared with 69.6 ± 28.1; t equals 2.50; P < 0.01). We tested for significance of interactions between minority status of physicians and their specialties and between minority status and practice volume. We found that the minority status of physicians was significantly associated with participatory decision-making style, whereas the interactions between minority status and practice volume and between minority status and specialty were not.

In a physician-level analysis using observed scores minus expected scores based on characteristics of each physician's Medical Outcomes Study patient panel, the physician characteristics we considered explained 52.6% of the reliable variation in participatory decision-making style. Neither physician age nor physician sex, however, was statistically significantly related to participatory decision-making style scores.

Discussion

Top Methods Results Discussion Author & Article Info References

We found that participatory decision-making style is a valid and reliable indicator of the quality of interpersonal care, as supported by its positive association with patient satisfaction and its negative relation to provider loyalty. In an earlier study of 263 patients with rheumatoid arthritis [13], we found a positive, statistically significant correlation between the three-item participatory decision-making style measure and a conversation-based measure of physicians' involvement of patients in treatment decisions derived from audiotapes (r = 0.24; P < 0.01).

After adjustment for patient characteristics, our findings suggest that lower practice volume, previous primary care or interviewing skills training, satisfaction with personal autonomy, white race, and specialty are all associated with higher participatory decision-making style ratings. Recent evidence suggests that as the pressures to contain costs increase, physicians respond by increasing the volume of their practices, with a corresponding decrease in time spent per patient [14, 15]. Medical Outcomes Study physicians in busy, high-volume practices—regardless of type of practice organization—were rated as less participatory than were those in lower-volume practices.

Busy practices with shorter office visits have been empirically linked with less effective patient behavior during those visits. Patients were shown to be less effective in information seeking during visits lasting less than 18 minutes [16]. More time with the physician has been associated with more favorable ratings of the encounter by the patient [17], especially if that time is spent in providing explanations of and rationales for treatment [18]. Our data suggest that more time with patients may be required to present and discuss treatment options and arrive at mutually acceptable treatment plans. If such participatory care has important consequences for health outcomes [1-5, 7], then cost-containment responses that increase physician practice volume may be compromising both the interpersonal quality of care and the effectiveness of that care.

Differences between specialties in cost and quality of care have been the subject of several recent empiric studies [9, 19, 24]. Although office visits to general internists are longer than visits to family practice physicians (18.4 minutes compared with 13.0 minutes) [20], few studies have explored differences between specialties in interpersonal care, and none of these have taken patient differences into account [20, 25, 26]. Among the specialties that we studied, cardiologists had the lowest participatory decision-making style ratings. Cardiologists have been rated higher than internists or family practice physicians on questionnaire-based assessments of appropriate technical care [27]. However, little is known about the variations in interpersonal style of care among specialists (such as propensity to involve patients in care) that may also have important consequences for patient outcomes.

Residents trained in primary care-track or interviewing skills training programs [28-30] have been shown to be less dominant during the medical interview, asking questions that are more likely to elicit accurate information from patients and receiving better ratings of communication skills from simulated patients [31]. Physicians with less dominant communication styles have been shown to have higher patient satisfaction ratings [25, 32, 33]. Few, if any, of the evaluations of these training programs include practicing physicians [30]. Our data suggest that these programs may have long-term effects, influencing physicians' approaches to patients some 10 years after training. Because these results persisted after adjustment for patient and physician characteristics, they provide compelling evidence that primary care programs or training in interviewing or interpersonal skills produce measurable and substantive differences in the overall quality of interpersonal care.

Physicians who were satisfied with personal autonomy in their professional situations, independent of practice organization, were rated as more participatory than physicians who were not satisfied. If a sense of personal autonomy is related to a more participatory style, then those physicians who view themselves as more able to control their practice environment may show a more flexible style with their patients. If a participatory style is related to patients' ratings of physicians' interpersonal care, patient follow-through on treatment recommendations, and other positive health behaviors [25, 32-34], then practice situations that foster a sense of autonomy among physician constituencies may enhance the quality of interpersonal care those physicians provide. Cost-containment strategies that limit or are perceived to limit practice autonomy may therefore have unintended consequences on quality of care.

Nonwhite physicians were reported by their patients to be less participatory than white physicians, after adjustment for the greater practice volume of nonwhite physicians. If nonwhite physicians are responding to their patients' preferences, then the effects of a less participatory decision-making style on subsequent outcomes should differ for various subgroups of patients and physicians. Current efforts to explore cultural differences that are important to the quality of interpersonal and technical care may help show whether these style differences have meaningful implications for the outcomes of care [35-37].

Over the past decade, the notion that patients should have a greater say in treatment decisions has evolved from advocacy to the identification of specific elements of interpersonal care that are concrete and measurable, that can be incorporated into training, and that yield positive health outcomes. In previous studies, we and others [1-57, 38-44] have shown that when physicians are less conversationally controlling during office visits (asking fewer closed-ended questions, giving fewer directions, interrupting less frequently, and involving patients in treatment decisions), patients have better health outcomes. Data from this study suggest that giving patients choices about, control over, and responsibility for certain aspects of care have important implications for patient loyalty and satisfaction with care. Current studies exploring the links between participatory decision-making style and patient out comes will help support this issue [45, 46].

Our study has some important limitations. First, we measured participatory decision-making style at a single point in time in relation to a specific office visit. However, we believe that physician style is a reproducible attribute, and this is supported by the relatively high level of agreement among the patients of each Medical Outcomes Study physician, most of whom had known that physician for several years.

Second, although we adjusted for case mix by using the presence of common comorbid conditions [47] and patients' ratings of health status, unmeasured case mix may have contributed to the association we observed between participatory decision-making style and specialty. Third, the Medical Outcomes Study physician sample was not random and did not represent all possible physician variables (such as specialty, ethnicity, and geographic region). Some comparison groups contained few physicians, which may limit the generalizability of our findings. However, the observed associations were independent of other physician characteristics, patient characteristics, site of care, and type of practice organization. Further empiric studies among different physician groups are needed, but the generalizability of our findings is plausible.

Finally, because our study was cross-sectional and only associational, no causal inferences can be drawn. For example, whether primary care-track or interviewing skills training causes behavioral differences or whether the physicians who choose such training differ in ways important to style must be determined by future research.

Recent restructuring of the practice of medical care to accommodate cost-containment initiatives may impinge on physicians' intentions or abilities to interact effectively with patients. Physicians in busy practice situations, despite a predisposition for a more participatory style, may not spend enough time with each patient to involve them effectively in management decisions. Since 1986, when this study was done, practice volume in managed care settings have tripled in volume over that observed in the fee-for-service sector. Changes of that magnitude would yield a six-point difference (100% of a standard deviation) in adjusted participatory decision-making style. These changes suggest that our data may underestimate rather than overestimate the consequences of variables such as practice volume on the quality of interpersonal care.

No amount of technically excellent care will produce optimal outcomes if patients are not actively engaged in managing disease, particularly chronic disease. If future research supports our results, then attention must be paid to factors supporting or compromising physician ability to provide high-quality interpersonal care in order to maintain patient outcomes. Such support for interpersonal care might be accomplished by the following: 1) increasing administrator awareness of the need to balance efficiencies gained by increasing practice volume with inefficiencies in suboptimal patient outcomes, provider switching, and physician dissatisfaction; 2) compensating physicians for interpersonal care through strategies such as reimbursement for "cognitive services"; and 3) furthering practice-based outcomes research to identify the "floor" for visit length below which quality of care is at risk, especially for vulnerable patient groups, such as the elderly or others for whom the interpersonal care process may require additional time.

Ms. Gandek and Drs. Rogers and Ware: New England Medical Center Hospitals, The Health Institute, 750 Washington Street, Box 345, Boston, MA 02111.