Previous studies found weak preferences for physician-patient gender synchronicity when presenting complaints were not gender-specific.1,2 Preferences for same-gender physicians became much stronger when subjects sought help for intimate health problems, including gynecologic and obstetric care.1,2
Although not linked directly to those preferences, communication studies of primary care visits found broad differences between physicians of different genders.3,4 Female physicians engaged in more partnership building, emotionally focused talk, positive talk, and psychosocial exchange than male physicians. Both male and female patients of female physicians also disclosed more biomedical and psychosocial information and were more positive in their talk than patients of male physicians.3,4
Communication behaviors associated with female physicians were generally valued by subjects and predictive of positive outcomes, including satisfaction, recall of medical information, compliance with medical recommendations,5 and health status improvements.6 The literature directly relating physician gender and patient satisfaction is mixed; some studies found higher satisfaction with female physicians7 and others the opposite.8,9
Although the dynamics of physician gender preferences and their consequences for communication and satisfaction are important to obstetric practices, none of the studies involved obstetricians.10 That was not surprising; most communication studies were conducted in primary care, and it is only recently that sufficient numbers of female physicians made such research practical.
As part of a larger study focused on prenatal genetic testing and counseling,11 we investigated the influence of physician gender on communication in obstetric visits. The present study of the routine process of care during first obstetric visit, based on audiotape analysis of male and female obstetricians, determined whether there were gender differences in style and content of obstetrician communication during prenatal visits and whether patient satisfaction with prenatal care was associated with obstetrician gender, even when controlling for communication and subject sociodemographic variables.
Materials and Methods
Physicians were recruited from the American Medical Association masterfile, which included board-certified or board-eligible obstetricians who graduated from medical school after 1970. A letter was sent to obstetricians explaining that the purpose of the study was to describe the content and process of early prenatal care for women of advanced maternal age. Participation would involve allowing the study team to solicit participation of women meeting eligibility criteria (33 years or older at delivery and less than 18 weeks' gestation at first prenatal visit) and allowing audiotaping of each first visit with up to ten consenting eligible patients. A postcard reminder was sent to nonrespondents 1 week after initial mailing, and a second mailing was sent 1 month later.
The age-eligibility criterion was later changed to include all women over 21 years old because of slow enrollment.
The subject recruitment protocol required a staff member from each physician's office to telephone eligible women before their first scheduled appointment to outline the study and obtain permission to have study investigators contact them. If permission was given, study staff telephoned women to briefly explain the protocol and obtain tentative agreement to participate. More details were discussed and informed consent was obtained in face-to-face meetings at physicians' offices immediately before scheduled visits. The visits were recorded on audiotape, and patients completed questionnaires immediately after the visits. The study method was approved by the Johns Hopkins Institutional Review Board; all participants (subjects and physicians) gave informed consent and were aware of the audiotaping.
Visits were audiotaped in physicians' offices during 15 months beginning in June 1994. Audiotapes were analyzed using the Roter Interaction Analysis System, which codes each complete thought, expressed as a statement or phrase, by patient or physician, into mutually exclusive and exhaustive categories. The categories relate broadly to the task of the visit and the social, emotional, and facilitative functions of the visit. Coders rated the content categories and the global emotional tone of physician and the patient (separately) during the visits in eight categories (anger, anxiety, dominance, interest, friendliness, responsiveness, sympathy, and hurriedness) using a five-point scale.
For the current study, the system was modified from previous applications by adding two categories tailored to prenatal discussion: counseling in preparation for and social adjustment to pregnancy (eg, birth and infant care classes, selection of pediatrician, circumcision, and postpartum support); and physical adjustment to pregnancy (eg, health promotion topics including diet, exercise, substance abuse, proper use of seat belts, and exposure to solvents, etc).
Table 1 presents the individual code categories within each of the content groupings and examples of dialogue from each coding category. Coding of audiotapes was done directly without transcription and averaged about 1½ times the length of the visit. The coding system showed good intercoder reliability, with an average reliability coefficient of .93 (range .51–1.00) for physician categories and .90 (range .67–1.00) for patient categories based on double coding of a 10% random sample.
The ratio of physician-to-patient talk, a measure designed to show the balance of dialogue in each visit, was calculated by dividing the frequency of all physician statements by all patient statements made during the visit. A patient communication control score was also calculated, relating patient control of information (patients' questions and physicians' answers and counseling, both biomedical and psychosocial), to physician control (physicians' questions, orientations, and patients' biomedical information). That score was used previously to characterize patient-centered interviews.12,13
Visit length in seconds was derived from audiotapes and it excluded waiting time.
Immediately after medical visits, patients completed questionnaires that included sociodemographics, emotional health-status questions, and a 20-item satisfaction questionnaire derived from previous studies.4,8 Factor analysis identified the following three characteristics of patient satisfaction: emotional responsiveness (seven items; Cronbach alpha = .66), examination-specific satisfaction (three items, Cronbach alpha = 56); and informational partnership (eight items, Cronbach alpha = .77). A one-item satisfaction question also was asked, using the same response set. Because of high positive skew, the satisfaction scores were dichotomized to reflect very satisfied versus other responses.
Associations between physician gender and communication were found through bivariate and multivariate analyses using the generalized estimating equation to adjust for the effect of multiple subjects clustered within each physician category.14 Communication was measured as frequency of statements and as a proportion of statements in a given category relative to all physician statements made during the visit. Because the generalized estimating equation calculates within-physician clustering effect, physicians with only one patient were excluded from analysis, which affected five physicians and resulted in elimination of one female and four male physicians and their five patients from analysis. The final number of physicians for the generalized estimating equation analysis was nine female (with 55 patients) and seven male physicians (with 27 patients).
Three hundred forty-nine physicians were contacted by letter for participation in the study, of which 184 (53%) responded, 129 (70%) declined, and 55 (30%) expressed willingness to participate. Twenty-one (38%) of the physicians who expressed initial interest in the study (6% of those on our mailing list) ultimately participated. Many physicians deferred a final decision to their office managers after indicating a willingness to participate. Most of those managers declined participation because of insufficient numbers of eligible women or the time demands of the recruitment protocol.
Half the physicians recruited to the study were male (n = 11) and half were female (n = 10). The average age of physicians was 43 years (range 32–49 years) with an average of 15 years of experience since graduation from medical school (range 7–24 years). There were no gender differences in age or time since graduation.
Table 2 shows the characteristics of study subjects. Patients of male and female obstetricians did not differ significantly by any of the characteristics.
Because of our interest in prenatal genetic testing and counseling,11 women of advanced maternal age were targeted. More than half the women participating in the study were 33 years old or older, white, well educated, with some form of medical insurance. Most women had at least one child and had not previously received care from study obstetricians.
We could not determine the total number of women contacted by each physician's office staff or the number of women who refused to have investigators contact them, which was anecdotally reported to be rare. When contacted by study staff, patient refusal rate was 5%.
Individual women were included in the study only once each. The 11 male physicians in the study averaged 2.8 patients (range 1–7 patients); the ten female physicians averaged 5.6 patients (range 1–13 patients).
Visits with male obstetricians averaged 5.5 minutes longer than those with female obstetricians, male physicians averaging 26.0 minutes (range 5.2–42.8 minutes) and females averaging a significantly shorter visits of 20.5 minutes (range 5.6–44.9 minutes, P < .05). An average of 5 minutes more was spent with women of advanced (35–43 years) or near-advanced (33–34 years) maternal age compared with younger women (21–32 years) (31 minutes for male and 26 minutes for female obstetricians).
Male obstetricians not only spent more time with their patients but more time was spent on dialogue. Male obstetricians made about 30% more statements than female obstetricians during a typical visit (377 versus 290 statements, P < .05). Male physicians and their patients talked more during visits than female obstetricians and their patients. The ratio of physician-to-patient talk indicates that male and female physicians contributed to medical dialogue at about the same rate (1.59:1 for female and 1.67:1 for male obstetricians), indicating that the greater amount of male physician talk during visits was proportionately matched by increases in patient talk.
As shown in Table 3, most of the dialogue of visits related to biomedical information-giving in two primary areas, information about the patient's medical condition and relevant history, and discussion about treatment, including medication, supplements, or tests. Counseling statements, defined here as persuasive communications intended to change or modify behavior, were less frequent and tended to be related to treatment recommendations.
Discussions of psychosocial and lifestyle topics were only about one quarter as frequent as biomedical exchanges. The highest frequency of psychosocial topics related to physical adjustment to pregnancy. Discussions about social and emotional adjustments to pregnancy were relatively infrequent.
Data gathering communication by physicians was most often through close-ended, biomedical questions. Less than 10% of all questions asked were open-ended. There were no statistically significant differences in any of those categories attributable to physician gender.
Socioemotional exchanges comprised a notable portion of dialogue, primarily through high levels of agreement. Physician gender effects were evident in this category, with female physicians showing higher proportionate levels of agreement and laughter. Female physicians also showed more frequent and proportionate disagreements; however, they were rare, averaging only once every two visits.
Categories of talk with explicit emotional content were infrequent, with the exception of reassurance. Physician gender differences were evident by significantly more statements of concern and partnership made by male physicians. Partnership-building categories of talk also showed some physician gender effects. Male physicians used significantly more checks, which included paraphrasing and interpretations, than females. Male physicians made significantly more orientation statements than females to help patients anticipate procedures and transitions during visits.
There were no indications that physician gender affected patients' talk during visits or coders' global ratings of physicians' or patients' affective tones.
Subjects were generally satisfied with their care. For overall satisfaction and satisfaction with physical examinations, 51 (74%) and 50 (72%) patients, respectively, rated their care as very satisfying; for satisfaction with emotional responsiveness, 38 (55%) women used the highest ratings; for satisfaction with informational partnership, 23 (38%) used the highest ratings.
After we accounted for communication and sociode-mographic variables, female physician gender predicted two of four measures of patient satisfaction, shown in Table 4.
The significant positive predictors of the one-item satisfaction measure included physicians' emotional talk and open-ended questions. Coders' global judgment of patients' negative affect was a negative predictor of satisfaction. Patients with lower income and patients with a positive self-rating of emotional state were more satisfied than other patients. When those variables were controlled for, female physician gender was not statistically significant in predicting global satisfaction.
Patient satisfaction with physicians' emotional responsiveness was predicted by patient communication control score and by the amount of psychosocial and biomedical counseling given to the patient. Patient race (minority women were more satisfied) and whether the patient received care from that physician previously were significant predictors of emotional satisfaction. After controlling for those other factors, female physician gender entered the model at a significant level (odds ratio 21.7).
Satisfaction with informational partnership was predicted by amount of biomedical information given by the physician, if the patient had received care from that physician previously, and patients' positive self-rating of emotional state. After controlling for those factors, female physician gender was a positive significant predictor.
Satisfaction with the physical examination was positively predicted by physician counseling on psychosocial issues and patients' positive self-rating of their emotional states. Physician disagreements and lower coders' global judgment of positive physician affect during visits were also positively associated with that aspect of satisfaction. Physician gender did not contribute significantly to the model.
The first obstetric visit presents an opportunity to establish therapeutic rapport and explore medical, social, and emotional adjustments to a pregnancy. However, we found that the focus of those visits was largely biomedical. Any discussions beyond strictly biomedical issues focused on health promotion counseling and physical adjustment to pregnancy. Broader discussions of social and emotional issues were infrequent. Although such discussions might take place at subsequent visits, the first visit is especially important in setting women's expectations for what can be considered appropriate topics of discussion.10
Male obstetricians conducted longer visits and engaged in more dialogue than female obstetricians. They were more likely to check that they understood the patient through paraphrasing and interpretation and to use orientations to direct the patient through the visit. Male physicians expressed more concern and partnership than female physicians. An exception to that pattern of findings was the higher proportionate level of emotionally charged talk, both agreements and disagreements, with female physicians. Consistent with the literature on communication and gender effects, those higher rates might reflect more intense affective engagement between female obstetricians and their patients compared with male obstetricians.15 Most categories of emotional and psychosocial exchange did not differ by physician gender, as would be expected considering the bulk of research in that area.3,4,15
Satisfaction findings were generally sensitive to task-focused and socioemotional elements of communication. Those correlations were mostly as expected; more emotional talk, psychosocial counseling, biomedical information-giving and counseling, and more open questions predicted several aspects of satisfaction. Coders' global judgments of the patients' affective tone expressed during visits provided additional concurrent validity to satisfaction measure. Women who sounded less negative during visits were more satisfied than others.
Several patient attributes were related to satisfaction. Minorities and lower income patients expressed higher satisfaction than others, which might be due to lower expectations for care or a less consumerist appraisal of the health care system and physicians. Women with a prior relationship with their physician expressed higher satisfaction, as expected. Patients' self-rating of emotional health was also related to satisfaction; healthier women were more satisfied than sick or distressed women, which was consistent with other studies.16
We found that in addition to the explanatory power of particular communication and patient variables, physician gender predicted emotional and informational satisfaction. The relationship of gender to emotional responsiveness was especially strong and might indicate sensitivity to gender expectations; women are generally perceived as more emotionally communicative than men.15
Male obstetricians in this study might have tried especially hard to please their patients, the most striking evidence of which was longer visit lengths and more talk. That finding was surprising because the National Ambulatory Medical Care Survey report in 1980 found only slight differences in length of visit in obstetrics and gynecology attributable to physician gender and a general trend for longer visits with female compared with male physicians.17
Increased visit length is not without considerable cost to physicians, but it is usually associated with payoffs in patient satisfaction, if only weakly.5 That was not the case here; none of the satisfaction dimensions measured in this study were related to length of visit.
As in any observational study, methodologic limitations should be considered. Physician participants were willing to allow researchers to audiotape their visits and provide full access to their patients and practices. Those physicians represented only 6% of all those contacted and might not be representative of obstetricians in general. The voluntary nature of participation demands self-selection and was characteristic of virtually all studies of this kind. Generalizability concerns regarding patient participants were less serious because the refusal rate was low.
Another study limitation relates to interpretation of observed behavior, raising the possibility of performance bias. Several studies specifically addressing that issue concluded that the presence of a tape recorder produces minimal reactive effects and does not affect doctor-patient communication in any systematic manner.18 Even if audiotaping inspired best behavior in our study, it is unlikely that the behavior was systematically biased in a way that would jeopardize interpretation of study findings.
Physician gender appeared to be an independent predictor of patient satisfaction, which for managed care might indicate that male obstetricians are at an unfair disadvantage if their satisfaction ratings are being contrasted with those of their female colleagues.