Abstract
Background: Although differences between Internal Medicine (IM) and Family Practice (FP) physicians have been examined in terms of care outcomes and cost, there have been few studies of specialty differences in physician-patient communication.
Methods: In 1995, 277 clinical encounters with 29 full-time, community-based FP physicians and 287 clinical encounters with 30 full-time, community based IM physicians were audiotaped. Communication was evaluated with the Roter Interaction Analysis System to reflect data gathering, patient education and counseling, rapport building, partnership building, verbal dominance, and patient-centeredness. Patient satisfaction was measured with an exit questionnaire.
Results: IM clinicians ask more biomedical questions (P = .02). FP clinicians engage in more psychosocial discussion (P = .02) and tend to engage in more emotionally supportive exchanges such as empathy and reassurance (P = .06). Significant interaction effects show differential treatment of patient subgroups by specialty; FP physicians were more verbally dominant with female patients (P < .01) and more patient-centered in their communication style with minority patients (P = .03). Although patient satisfaction was similar for IM and FP, satisfaction was more closely linked to measures of rapport and patient-centeredness for patients of FP physicians than for patients of internists.
Conclusions: The current work adds insight into FP and IM differences in both physician-patient communication and predictors of patient satisfaction.
Multiple reports have detailed differences in the process and cost of care provided by Internal Medicine (IM) and Family Practice (FP) clinicians.1–24 The origin of these specialty differences may stem in part from curricular priorities exhibited in residency education. FP residency programs devote 3 times as many formal training hours to psychosocial issues and interviewing skills as IM residency programs.25
The communication patterns of IM and FP have been compared in 2 reports both describing a single set of resident physicians at the University of California Davis Medical Center. Bertakis et al26,27 used the Davis Observation Code (DOC) to analyze initial and subsequent clinical encounters for patients randomly assigned to the FP or IM resident physician clinics. Both reports conclude that the IM residents spent more time collecting information and discussing treatments and that the FP residents focused more on emotional counseling than did the IM residents. Physician specialty did not predict patients’ self-reported health status or satisfaction.
Based on these observations26,27 and our own prior work, hypotheses were developed regarding the communication practices of IM and FP clinicians in relation to the 4 functions of the medical interview: (1) data gathering, (2) patient education and counseling, (3) rapport building, and (4) partnership building.28,29
We hypothesized that: (1) IM physicians would conduct visits with more data gathering, (2) FP physicians would undertake more patient education and counseling, (3) FP physicians would exhibit more rapport building, and (4) FP clinicians would exhibit more partnership building. In addition, we hypothesized that: (5) IM clinicians would be more verbally dominant in their patient exchanges and (6) FP clinicians would score higher on a summary measure of patient-centeredness. Finally, because satisfaction measures are typically correlated with patient-centered communication elements, we hypothesized that patient satisfaction would be higher for FP physicians.
Subjects and Methods
To evaluate these hypotheses, we conducted a secondary data analysis of 564 audiotaped patient encounters conducted in 1995 with 30 IM and 29 FP community-based clinicians. Audiotapes of these encounters were collected as part of a randomized trial of communication skills training.30
Physicians
Full-time primary care physicians in the greater Baltimore metropolitan area were recruited from local medical society mailing lists to participate in a study to evaluate a continuing medical education (CME) program. Eighty-eight physicians (16% of those contacted) agreed to all aspects of the study and 69 of these physicians fully participated. Ten of these physicians were not board-eligible and were excluded from analysis. Of the remaining 59 physicians, 29 were identified as Family Physicians and 30 were identified as Internists.
Informed consent was obtained at enrollment of physicians in the study. Physicians were located in solo, small group, and HMO settings. Female physicians accounted for 10% of the sample. Physicians contributed an average of 9.6 (range, 6 to 11) audiotapes. IM physicians were older with a greater number of years in practice. Physician characteristics are summarized in Table 1.
Physician Characteristics
Patients
Patients were recruited immediately preceding their medical appointments; informed consent was obtained at the time of recruitment. Seventy-two percent of patients who were approached gave their consent. Patients completed short pre- and postvisit health-status questionnaires that included self-ratings of physical and emotional health.31 In addition, at the conclusion of their office visits, patients completed a 16-item satisfaction questionnaire with adequate internal reliability (α = 0.87).29
Differences between the IM and FP patients were determined for patient age, income, and work status; the IM patients were older, less wealthy, and reported lower levels of employment. Patient characteristics are summarized in Table 2.
Patient Characteristics
Study Variables
Audiotapes were analyzed using the Roter Interaction Analysis System (RIAS), a widely used quantitative approach to audiotape coding.32 The RIAS catalogues each complete thought expressed by the patient or physician into mutually exclusive and exhaustive categories. The categories of the RIAS taxonomy correspond to the biomedical and socioemotional functions of the visit.29,33
The RIAS codes were divided into composite summary measures to evaluate our specified hypotheses. The first 4 composites regarding data gathering, patient education and counseling, rapport building, and partnership building are the simple sum of their constituent code frequencies, as depicted in Table 3.
The Roter Interactional Analysis System (RIAS): Frequency of Four Code Groupings According to Physician Specialty*
Verbal dominance was defined as the sum of all physician statements divided by the sum of all patient statements, as has previously been described.34 Patient-centeredness was defined as the sum of all codes relating to socioemotional handling and the psychosocial agenda of the patient divided by the sum of all codes related to furthering the biomedical agenda.35 The numerator for this equation includes all patient-centered RIAS codes: all physician and patient comments and questions relating to psychosocial and lifestyle-related topics, all physician and patient statements of partnership building, social talk, positive talk, and statements that respond to emotions. The denominator for this equation includes all doctor-centered RIAS codes: all physician and patient biomedical question asking and information giving and physician counseling on biomedical topics. As such, a value greater than 1 denotes a more psychosocially oriented encounter and a value less than 1 denotes a more biomedically oriented encounter.
The audiotapes were coded by 3 persons trained in the RIAS system. Reliability of coding was assessed by double-coding a 10% random sample of audiotapes drawn throughout the coding period. The average intercoder reliability was .73 and ranged between .62 and .98 for individual categories. In addition to the discrete category coding, coders also rated the emotional tone of the visits on a 6-point Likert scale for 5 affective domains: anger, dominance, interest, responsiveness, and friendliness. Coder agreement on the affect ratings within 1 point ranged from .63 to .84.
Data Analysis
The communication patterns of IM and FP physicians were compared by contrasting the mean counts of their behaviors within the 4 functions of the visit (data gathering, patient education and counseling, rapport building, and partnership building), and in verbal dominance and a summary score of patient-centeredness. These same analyses were conducted with encounter duration as a single control variable.
Next, we performed multivariate analyses for data gathering, patient education and counseling, rapport building, partnership building, verbal dominance, and patient-centeredness. As physicians contributed a varied number of patient encounters, general estimating equation (GEE) techniques for panel data were used to cluster data by physician.36 Patient and physician sociodemographic characteristics were entered for adjustment to take into account underlying differences in these characteristics. Control variables, including physician age, experimental group status, years in practice, patient age, patient income, patient sex, patient ethnicity, history of prior visits, Generalized Health Questionnaire (GHQ) score, and duration of encounter were specified a priori and additional characteristics and interactions were explored to improve model fit.
Statistical analyses were conducted using STATA version 7.0 (Stata Corporation, College Station, TX). The authors had full access to all the data in the study, have no financial or personal conflict of interest, and accept responsibility for the integrity of the data and analysis.
Results
RIAS Communication Analysis
Table 3 exhibits the frequency of interaction categories by physician specialty. Interaction categories that showed specialty differences in bivariate analysis included higher levels of closed-ended biomedical questions and bids for clarification for Internists, and higher levels of psychosocial information giving for family physicians. A trend suggesting higher levels of empathy for family practitioners was also evident (P = .07). When adjusted for encounter duration, higher levels of closed-ended psychosocial questions by internists were also evident.
Inclusion of additional potential confounding variables related to the original study intervention (ie, experimental group membership and the patient’s GHQ score) as well as other patient and physician characteristics (ie, being an established patient, physician age, patient income, patient ethnicity, and patient age) in an expanded model, showed persistent specialty differences favoring internists in higher levels of data gathering and family practitioners in higher levels of psychosocial information giving. In addition, family practitioners showed higher levels of laughter and trends indicating higher levels of empathy and reassurance (P = .06, respectively). No differences were evident in the partnership building categories.
Verbal Dominance and Patient-Centeredness
As seen in Table 4, FP physicians exhibited significantly more verbal dominance than IM physicians. After controlling for potential covariates, patient gender emerged as a significant interaction variable; between-specialty differences in verbal dominance were evident only with female patients. Similarly, after controlling for covariates, patient ethnicity emerged as a significant interaction variable such that between-specialty differences in patient-centeredness were evident only for African American patients.
Determinants of Verbal Dominance and Patient-Centeredness
Global Assessment
Global affect ratings by coders for physicians revealed a higher rating of friendliness for FP (IM = 4.0, FP = 4.3; P = .05) but no differences in other affective ratings. No differences were found in the raters’ global affect ratings of patients.
Patient Satisfaction
No specialty distinctions were evident in patient satisfaction ratings (P = .91). For both IM and FP, satisfaction was correlated with rapport building and psychosocial communication categories. However, this association was substantially stronger for patients of FP than for IM (Table 5); the correlations between patient satisfaction and both psychosocial and rapport building communication were more than doubled in magnitude for family physicians than internists in each case. We also found substantially higher correlations between psychosocial talk and satisfaction for female compared with male patients (0.20 vs 0.03), regardless of physician specialty. For the summary score of patient-centeredness, patient satisfaction showed a significant correlation only for patients of family practitioners.
Correlation between Patient-Centeredness and Patient Satisfaction*
Discussion
Consistent with previous studies with medical residents,26,27 community-based IM physicians engage in more data gathering with their patients, and FP clinicians provide more psychosocial information and tend to engage in more emotionally supportive exchanges such as empathy and reassurance with their patients. We also found some evidence that FP clinicians may relate to particular subgroups of patients in a different manner than their IM colleagues; FP physicians were more verbally dominant with their female patients and more patient-centered with African American patients than were Internists.
Further exploration of our data showed that the increased verbal dominance exhibited by FP physicians with their female patients was a result of a higher level of physician psychosocial talk. In light of the substantially higher correlation between psychosocial talk and satisfaction for female compared with male patients, FP clinicians may be responding to patient cues that encourage this type of communication.
Evaluation of the finding of increased patient-centeredness directed toward African American patients by FP clinicians reveals especially high levels of rapport building communication. This may reflect an attempt by FP clinicians to ameliorate the more negative experiences of African American patients with the medical system compared with their white counterparts. For instance, the Commonwealth Survey found that 15% of African American patients believed that they would receive better health care if they were of a different race or ethnicity; the survey also found that African American patients were almost twice as likely as their white counterparts (16% vs. 9%) to report treatment with disrespect during a recent health care visit.37
Although a statistical trend existed for an increased number of empathic and reassurance statements made by FP physicians in this study, the actual mean number of empathic statements made per encounter was very small (0.5 and 0.3 for FP and IM, respectively; P = .06). It is possible that statements of empathy represented critical incidents of affective bonding, and even modest differences in the use of this low incidence code represent clinically relevant distinctions in communication. Although the frequency of reassurance statements was also higher for FP than IM (5.6 and 4.7, respectively, P = .06) the difference was also quite small. It is notable that the frequency of these behaviors and several other powerful tools of rapport and partnership building was low for both IM and FP physicians.
Patient satisfaction was similar for IM and FP encounters. However, the pattern of association between patient satisfaction and communication categories revealed an unanticipated specialty difference. The satisfaction of FP patients was more closely linked to rapport building, psychosocial exchange, and patient-centeredness than they were for patients of IM clinicians. It is possible that patients who select FP clinicians may do so because they believe and expect FP to be more psychosocial and emotionally supportive and less likely to exhibit a biomedically intensive communication style than Internists. In response, the physicians may indeed exhibit those qualities of exchange that are expected.38 In this light, some measure of the difference between the 2 specialties may stem from cultural differences both in specialty training and in the background and expectations of the patients who frequent these clinicians.
This study has certain limitations. First, as a secondary data analysis, this study was not designed to evaluate variability by specialty, but to assess a communication skills training intervention. To control for the possibility of unmeasured differential impact of the intervention on physician specialty, experimental group status was included as a covariate in the expanded model. Second, physicians who had agreed to participate in the original study could have been different from those who declined. External validity is a common challenge in studies that demand time commitments from participating physicians, as was true for this study, and the volunteer rate was low (16%). However, the participation rate was about the same for both FP and IM physicians, and any selection bias should be equivalent in both groups. Third, these data are derived from encounters with a small sample of female and minority group clinicians. Consequently, physician gender and ethnicity could not reliably be used in analyses. Because of the small number of physicians in this analysis, only one physician variable (age) was included in expanded models. Fourth, because these data were collected in 1995, it is important to note that communication patterns described above may have shifted during the interim.
Conclusions
Since the specialty of FP was first recognized, in 1969, differences between the 2 fields have diminished in certain respects. Overall, because of the success of FP clinicians and tremendous shifts in health care systems, case-mix differentials have narrowed and will probably continue to diminish.12,23,39–41 However, Family Practice and general Internal Medicine remain distinct specialties with their own philosophies and curricula.42 These data exhibit significant differences in the way IM and FP physicians communicate and provide the first evidence from encounters with board-certified or -eligible physicians describing the role of specialty in communication patterns and corroborate prior research done with resident physicians. We do not know whether the communication differences evident here were caused by training and curricular emphasis of the specialties or the self-selection of physicians to specialties that are consistent with particular interests and style preferences. Certainly, patient self-selection may also play a role in explaining the specialty differences in predictors of patient satisfaction. Further studies are required to evaluate the health impact of these communication differences and to explore the interaction between specialty and patient and physician identity characteristics, such as ethnicity and gender.
Categories of Roter Interaction Analysis System (RIAS)
Notes
MP-O was supported by grant T32-HL07180 from the National Heart, Lung, and Blood Institute.
- Received for publication March 18, 2003.
- Revision received March 18, 2003.
References
Jump to section
Related Articles
Cited By...
- Communication quality in telephone triage conducted by general practitioners, nurses or physicians: a quasi-experimental study using the AQTT to assess audio-recorded telephone calls to out-of-hours primary care in Denmark
- Development of a tool for assessing quality of comprehensive care provided by community health workers in a community-based care programme in South Africa
- A profile of communication in primary care physician telephone consultations: application of the Roter Interaction Analysis System