To investigate the effect of patient gender on patient–physician communication in the process of recommendation for total knee arthroplasty (TKA).
To investigate the effect of patient gender on patient–physician communication in the process of recommendation for total knee arthroplasty (TKA).
Seventy-one physicians (38 family physicians and 33 orthopedic surgeons) in Ontario performed blinded assessments of 2 standardized patients (1 man and 1 woman) with moderate knee osteoarthritis and otherwise identical scenarios. Four surgeons did not consent to including their data. Standardized patients and accompanying mock family members recorded elements of informed decision making (IDM) as present/absent in the patient–physician discussion and rated the physicians' interpersonal skills.
Overall, the completeness of IDM was low, but was lower still for the woman. Only 57% (38 of 67) and 15% (10 of 67) of physicians discussed the nature of the decision and elicited the patient's preference while consulting with the man and woman, respectively. Even after adjusting for physicians' recommendations regarding TKA, when interacting with the woman, physicians included fewer IDM elements (adjusted mean difference in IDM score 1.2 [95% confidence interval (95% CI) 0.6, 1.8]; P < 0.001) and had poorer interpersonal skills (adjusted mean difference 14.1 [95% CI 9.0, 19.2]; P < 0.001) compared with their consultation with the man.
Physicians provided less medical information and less encouragement to participate in the decision to undergo TKA to a woman compared with a man, irrespective of their recommendation regarding TKA. Our findings suggest that in addition to directly influencing physicians' clinical decision making, gender bias may also influence physicians' interpersonal behavior.
Osteoarthritis is the leading cause of long-term disability () and is the most prevalent chronic condition among women (). For individuals with moderate to severe osteoarthritis who have not responded to medical therapy, total joint arthroplasty is considered the definitive treatment and can dramatically reduce joint pain and improve patients' function and quality of life (). Although age-adjusted rates of total joint arthroplasty are higher for women than for men (), based on a population-based epidemiologic survey, underuse of arthroplasty among willing and appropriate candidates is more than 3 times greater in women than in men (). The observed gender disparity in total joint arthroplasty at the population level makes this surgical procedure an appropriate model for the assessment of gender bias on clinical decision making. In prior research, we demonstrated that physicians were more likely to recommend total knee arthroplasty (TKA) to a man than to a woman, suggesting that physicians may be at least partially responsible for the gender disparity in the rates of use of total joint arthroplasty ().
Bias may also influence physicians' interpersonal behavior through their decision-making style and patient–physician communication (). A more participatory decision-making style and more effective patient–physician communication on the part of physicians have been shown to be positively associated with patient satisfaction, adherence to physicians' treatment recommendations, utilization, and improved health outcomes ([8-12]). As a result, a more patient-centered approach may help to remedy gender disparity (). The extent to which physicians involve their patients in decision making regarding TKA and whether this differs when the patient is a man or a woman is unknown. The purpose of this study was to investigate the effect of patient gender on patient–physician communication in the process of recommendation for TKA.
The data were collected as part of a larger study. Detailed methods are described elsewhere (). Seventy-one physicians (38 family physicians and 33 orthopedic surgeons) in Ontario, Canada were visited by 2 standardized patients between August 2003 and October 2005. Physicians were blind to the patients' status as standardized patients. Family physicians with open practices provided written informed consent prior to any standardized patient visits. Orthopedic surgeons who perform TKA had indicated their willingness to be visited by standardized patients in a prior mail survey (). After the standardized patients visited them, they were informed about the “temporary waiver” of consent granted by our ethics review committees; 4 surgeons did not consent to including their data. All physicians were told that they were participating in a study of clinical decision making, but were not told that the purpose of the study was to assess the effect of patients' gender on patient–physician interaction. Ethics approval was obtained from the University of Toronto and The Hospital for Sick Children.
A man and a woman living with moderate knee osteoarthritis, confirmed by physical examination and radiographs as being identical in terms of disease severity by our 2 orthopedic surgeon coinvestigators (HJK, NNM), were recruited as standardized patients. We used patients with osteoarthritis (rather than actors) to minimize the risk of unblinding physicians and to provide a realistic clinical presentation (). Patients were not obese and had only mild comorbidities considered normal for their age (e.g., controlled hypertension). The standardized patients were not informed of this study's purpose.
The standardized patients memorized identical scenarios for functional capacity, pain severity, and other treatment modalities () (see Supplementary Appendix A, available in the online version of this article at http://onlinelibrary.wiley.com/doi/10.1002/acr.21970/abstract). Their marital status was scripted as married, their socioeconomic status was scripted as middle class, and both had physically undemanding part-time jobs. Chronic knee pain was their chief symptom so that the physician encounter would focus exclusively on their osteoarthritis. The scenario was scripted to have exhausted usual nonoperative treatment.
Each standardized patient was accompanied to each clinic visit by a volunteer posing as his or her family member, forming a 2-member team. The gender of the mock family member matched that of the standardized patient. Their main function was to increase both the accuracy and reliability in recording the physicians' actions on the itemized postvisit checklist. To avoid influencing patient–physician interaction (), mock family members were trained to provide no medical information, to redirect any of the physicians' questions back to the standardized patient, and to respond in a manner that would substantiate the standardized patient's story regarding their pain or decreased function. Mock family members were only present in the examination room for the orthopedic surgeons.
In addition to the physicians' recommendation to refer for or perform TKA (yes/no; the results of which have been published elsewhere ), elements of informed decision making (IDM) ([17-20]) were recorded as being present (yes/no) during the consultation on the postvisit checklist. If the physician merely made mention of the element, the teams were instructed to record this as a “yes.” The IDM elements represent the decision-making style component of the physicians' interpersonal behavior. We chose the IDM scale ([17, 18, 20]) because it was previously used to evaluate the adequacy of orthopedic surgeons' IDM while consulting a standardized patient for possible total hip arthroplasty for osteoarthritis () as well as regular practice patients for a number of common surgical procedures ([18, 20]). The criteria for IDM included 42 items summarized as 7 elements () (Table 1 and Figure 1). For elements 2, 3, and 4, we specified the content of IDM that was particular to their scenario. For the discussion of the clinical issue, for example, items included how their arthritis developed, the surgery details, and what to expect during recovery. If even one of these items was present, that IDM element was considered to be present during the consultation. An IDM score (of 7) for each physician was derived from the sum of the IDM elements reported as present during the consultation by each team.
|Elements of IDM||Physicians including element in discussion, no. (%)||OR (95% CI)||P||Adjusted Pa|
|1. Discussion of the patient's role in decision making||42 (62.7)||17 (25.4)||4.9 (2.4, 10.0)||< 0.001||< 0.001|
|2. Discussion of the clinical issue or nature of the decision||57 (85.1)||31 (46.3)||6.6 (3.2, 13.8)||< 0.001||< 0.001|
|Informed patient about:|
|How the arthritis developed||56 (83.6)||30 (44.8)|
|Irreversibility of the arthritis||24 (35.8)||11 (16.4)|
|What part of the knee joint would be taken out||17 (25.4)||8 (11.9)|
|What the knee joint would be replaced with||16 (23.9)||9 (13.4)|
|How long the surgery would last||9 (13.4)||2 (3.0)|
|How long he/she would be in the hospital||20 (29.9)||6 (9.0)|
|How much pain he/she would have after surgery||3 (4.5)||5 (7.5)|
|How much help he/she would need after surgery||16 (23.9)||5 (7.5)|
|When he/she could start his/her usual activities||13 (19.4)||4 (6.0)|
|3. Discussion of the alternatives||48 (71.6)||66 (98.5)||0.04 (0.01, 0.27)||0.001||0.001|
|Tibial osteotomy||0 (0)||0 (0)|
|Arthroscopy||10 (14.9)||23 (34.3)|
|Rest or reduced activity||0 (0)||4 (6.0)|
|Assistive devices (e.g., cane, walker, brace)||13 (19.4)||9 (13.4)|
|Weight loss||0 (0)||8 (11.9)|
|Exercise||10 (14.9)||21 (31.3)|
|Patient education (e.g., pamphlet)||2 (3.0)||7 (10.5)|
|Social support||0 (0)||0 (0)|
|Non-narcotic analgesic (e.g., acetaminophen)||20 (29.9)||6 (9.0)|
|Narcotic analgesic||3 (4.5)||4 (6.0)|
|Nonsteroidal antiinflammatory drug||18 (26.9)||23 (34.3)|
|Chondroitin and glucosamine||5 (7.5)||13 (19.4)|
|Topical steroid cream||3 (4.5)||7 (10.5)|
|Increase existing pain medication||9 (13.4)||21 (31.3)|
|Joint injection with corticosteroids||12 (17.9)||27 (40.3)|
|Viscosupplementation (e.g., hylan G-F 20)||10 (14.9)||13 (19.4)|
|Referral to other health professionals|
|Rheumatologist||3 (4.5)||5 (7.5)|
|Physiotherapy||8 (11.9)||19 (28.4)|
|Nutritional or weight loss counseling||0 (0)||0 (0)|
|4. Discussion of the potential benefits and risks of the alternatives||36 (53.7)||15 (22.4)||4.0 (2.1, 7.6)||< 0.001||< 0.001|
|Discussed risks and benefits of surgery|
|Risk of blood clots||20 (29.9)||8 (11.9)|
|Risk of infection||21 (31.3)||7 (10.5)|
|Risk of heart attack or stroke||9 (13.4)||4 (6.0)|
|Risk of death||5 (7.5)||5 (7.5)|
|Risk of needing a revision knee arthroplasty||19 (28.4)||7 (10.5)|
|Benefit of increased mobility||30 (44.8)||10 (14.9)|
|Benefit of decreased pain||26 (38.8)||11 (16.4)|
|Discussed risks and benefits of alternatives|
|Risk of continued pain||24 (35.8)||8 (11.9)|
|Risk of decreased mobility||17 (25.4)||8 (11.9)|
|5. Discussion of the uncertainties associated with the decision||27 (40.3)||12 (17.9)||3.1 (1.7, 5.5)||0.004||0.024|
|6. Assessment of the patient's understanding of the decision||42 (62.7)||16 (23.9)||5.4 (2.8, 10.4)||< 0.001||< 0.001|
|7. Exploration of the patient's preference||41 (61.2)||15 (22.4)||5.5 (2.6, 11.4)||< 0.001||< 0.001|
Each team also rated the physicians' interpersonal skills as a measure of the patient–physician communication component of the physicians' interpersonal behavior (Table 2). Interpersonal skills were rated by responding to 11 items reflecting verbal communication skills (such as “using language you can understand when explaining your arthritis and treatment options”), paraverbal skills (such as “treating you like you're on the same level; not ‘talking down’ to you”), and nonverbal skills (such as “showing interest in you as a person”) (see Supplementary Appendix B, available in the online version of this article at http://onlinelibrary.wiley.com/doi/10.1002/acr.21970/abstract). Each of these 11 skills was rated on a 5-point scale from poor to excellent (range 0–4, where 0 = poor and 4 = excellent). The total score was calculated as the sum of the scores on each individual item, and then multiplied by a factor of 2.273 to obtain scores on a scale from 0–100. The teams also recorded the length of time they spent with the physician (in minutes).
|Mean (95% CI)||Mean difference (95% CI)||P||Mean (95% CI)||Mean difference (95% CI)||P|
|IDM score||1.8 (1.3, 2.3)||< 0.001||1.2 (0.6, 1.8)||< 0.001|
|Male patient||4.4 (3.9, 4.9)||4.1 (3.6, 4.5)|
|Female patient||2.6 (2.1, 3.0)||2.9 (2.4, 3.3)|
|Interpersonal skills score||19.4 (14.8, 23.9)||< 0.001||14.1 (9.0, 19.2)||< 0.001|
|Male patient||65.7 (61.5, 70.0)||63.1 (59.0, 67.1)|
|Female patient||46.4 (42.4, 50.3)||49.0 (45.0, 53.0)|
Statistical analyses were performed using SAS, version 9.0. Our study involved a binary response model, as each physician provided one response for the man and one response for the woman. We first used a global test for multivariate binary data, the simultaneous marginal homogeneity test (SMH; a score-type test statistic W0) (), to simultaneously test for evidence against the null hypothesis for all 7 IDM elements that the proportion of physicians including the IDM element during the consultation was the same for both the man and the woman. After finding a difference and rejecting the global null hypothesis at an alpha level of 0.05, we then proceeded with univariate analyses using binary logistic regression with a generalized estimating equation (GEE) approach to assess which of the IDM elements had a gender difference in proportions. The GEE method was preferred over conditional logistic regression because of its ability to account for the clustering effects of 2 patients per physician and any existing within-physician correlation of physicians' responses for the 2 patients (). An odds ratio (OR) with 95% confidence intervals (95% CIs) was calculated to estimate the strength of association between patient gender and inclusion of each IDM element during the consultation. PROC MULTTEST was used to obtain adjusted P values, adjusting for multiple testing.
Next, we calculated each physician's IDM score (of 7) and interpersonal skills score (of 100) during each consultation. We examined whether physicians' recommendations for TKA were associated with their IDM or interpersonal skills scores during the consultation for the combined data from the male and female standardized patients, and subsequently included it as a covariate in the analyses. For this GEE regression analysis, we used a normal distribution, an identity link function, and an independent correlation structure. We evaluated the effect of patient gender on physicians' IDM and interpersonal skills scores, adjusting for possible differences due to physicians' recommendations for TKA. After detecting an overall main effect for patient gender, we tested for an interaction for the prespecified subgroup of physician type. Then we evaluated the effect of physicians' recommendations for TKA on physicians' interpersonal behavior, stratified by gender. Finally, we evaluated the effect of patient gender on physicians' interpersonal behavior, stratified by physicians' recommendations for TKA. All tests were 2-tailed, with a significance level of α = 0.05.
Of the 71 physicians visited by standardized patients, 67 physicians (38 family physicians and 29 orthopedic surgeons) consented. Fifty-five physicians (82%) were men, and of the 12 female physicians, 9 were family physicians. The mean age of the physicians was 52 years (range 33–77 years) and the mean number of years in practice was 21 (range 2–49 years).
Overall, the completeness of IDM was low, but was lower still for the woman. Of 134 discussions, 1 did not include any IDM elements, while 26% (35 of 134) included only one of the elements. Although our study involved a complex clinical decision, only 9% (12 of 134) of discussions included all 7 IDM elements (13% [9 of 67] and 4% [3 of 67] of physicians during the consultation with the man and the woman, respectively). Only 57% (38 of 67) and 15% (10 of 67) of the physicians discussed the nature of the decision and elicited the patients' preference while consulting with the man and the woman, respectively; these 2 elements are considered to be the requisite elements for a basic clinical decision. Physicians included fewer IDM elements during their consultation with the woman compared with the man (SMH: W0 = 39.3, P < 0.001). As specified a priori, given the main overall effect, we proceeded with univariate analyses. With the exception of a discussion of alternatives to TKA, the odds of physicians including an IDM element were higher (ranging from 3–6 times higher) when consulting the man compared to the woman (P < 0.05) (Table 1). Among the alternatives to TKA, more physicians recommended arthroscopy, a cortisone shot, physiotherapy, weight loss, exercise, increasing existing pain medication, and glucosamine/chondroitin to the woman than to the man, whereas more physicians recommended acetaminophen to the man than to the woman. Physicians were less likely to discuss the clinical issue or nature of the decision with the woman compared with the man, including rarely discussing the 4 items regarding recovery postsurgery. Physicians seldom discussed the patient's role in decision making, explored whether the patient understood the decision, and elicited patient preferences when interacting with the woman versus the man (Figure 1).
Physicians' mean IDM and interpersonal skills scores were lower during consultations involving the woman compared with the man (P < 0.001). Even after adjusting for physicians' recommendations regarding TKA, when interacting with the woman, physicians included fewer IDM elements and had poorer interpersonal skills (P < 0.001) compared with their consultation with the man (Table 2). In analyses combining the physician assessments from both the male and female standardized patients, physicians recommending TKA had a higher mean IDM score (4.6 [95% CI 4.0, 5.1] versus 2.4 [95% CI 2.0, 2.8]; P < 0.001) and interpersonal skills score (66.2 [95% CI 61.4, 70.9] versus 45.9 [95% CI 42.0, 49.8]; P < 0.001) than physicians not recommending TKA. In addition, since orthopedic surgeons were more likely to recommend TKA (OR 3.1 [95% CI 1.6, 6.0], P = 0.001) than family physicians, they were among those considered to have a more participatory decision-making style. However, there was no significant patient gender by physician type interaction for either outcome (P > 0.20). Because few of the participating physicians were women, we could not examine the effect of physician gender on our findings.
Although physicians recommending TKA included more IDM elements and were rated as having higher interpersonal skills than physicians not recommending TKA while consulting both the male and female standardized patients (P < 0.01), they still included 1.6 fewer IDM elements (P = 0.001) and had poorer interpersonal skills (P < 0.001) with the woman compared to the man (Table 3).
|Evaluation category||Physicians recommending total knee arthroplasty||Physicians not recommending total knee arthroplasty||Mean difference (95% CI)||P|
|Mean (95% CI)a||Mean difference (95% CI)||P||Mean (95% CI)a||Mean difference (95% CI)||P|
|Male standardized patient|
|IDM score||5.1 (4.6, 5.6)||3.0 (2.1, 3.8)||2.1 (1.1, 3.1)||< 0.001|
|Interpersonal skills score||70.6 (66.2, 75.1)||55.7 (47.9, 63.5)||14.9 (6.0, 23.9)||0.001|
|Female standardized patient|
|IDM score||3.5 (2.6, 4.4)||2.1 (1.7, 2.5)||1.4 (0.3, 2.4)||0.009|
|Interpersonal skills score||57.1 (49.3, 64.9)||41.1 (37.4, 44.8)||16.0 (7.4, 24.6)||< 0.001|
|IDM score||1.6 (0.6, 2.5)||0.001||0.8 (−0.05, 1.74)b||0.065|
|Interpersonal skills score||13.5 (6.1, 20.9)||< 0.001||14.6 (6.6, 22.6)||< 0.001|
Physicians spent, on average, 19.6 minutes and 17.3 minutes with the male and female standardized patients, respectively (adjusted mean difference 2.3 [95% CI 0.1, 4.4]; P = 0.037). Orthopedic surgeons spent more time with patients than family physicians (adjusted mean difference 20.4 [95% CI 17.9, 22.9] versus 16.9 [95% CI 15.2, 18.7]; P = 0.029) and physicians recommending TKA spent more time with patients than physicians not recommending TKA (adjusted mean difference 20.8 [95% CI 18.5, 23.1] versus 16.1 [95% CI 14.5, 17.7]; P < 0.001); however, time spent with the male versus the female standardized patient did not differ for these physician subgroups. The median duration of the visits was 16 minutes (range 7–45 minutes). Patient–physician discussions with lower IDM scores were shorter than those with higher IDM scores (P < 0.001) (Table 4).
|IDM score||Duration of visit, minutes|
IDM has been defined as a meaningful dialog between patient and physician instead of a unidirectional, dutiful disclosure of alternatives, risks, and benefits by the physician to the patient (). With calls for more sharing of decision making (), the onus is on physicians to ensure that patients have all the information they need and that they understand their options, and then to help them to make an informed decision that corresponds with their preferences. This process is particularly important in making a decision about an elective surgical procedure such as TKA. We found that such dialog may be more problematic for women. Physicians in our study included fewer IDM elements and had poorer interpersonal skills during the consultation when the patient was a woman compared to a man. Physicians provided less medical information and less encouragement to participate in the decision to undergo TKA to the woman compared with the man, irrespective of their recommendation regarding TKA. In prior research, we demonstrated that physicians were more likely to recommend TKA to a man than to a woman (). Together, these results suggest that gender bias may be influencing both physicians' clinical decision making and physicians' interpersonal behavior, providing further support that physicians may contribute to the gender differences in the treatment of osteoarthritis ([5, 24, 25]).
This study adds to the increasing evidence that stereotyping based on patients' demographic characteristics such as race/ethnicity ([8, 9, 26, 27]), socioeconomic status (), and sometimes gender ([6, 28, 29]) affects physicians' treatment recommendations and physicians' interpersonal behavior during the clinical encounter, suggesting that physicians may contribute to disparities in health care utilization ([5, 30-34]). Although prior studies have shown that physicians provide less information and are less participatory with minority patients ([8, 9, 26]) and less educated and lower-income patients (), for gender, the data are equivocal.
Some studies have shown that women receive more medical information and have more participatory clinical encounters than men ([8, 9, 35]), whereas others have found the opposite () or clarified that women receive more information, since women ask more questions ([8, 35]). An important limitation of these observational studies is that some compared recordings of physicians' conversations with male and female patients whose chief symptom, presenting symptoms, and other medically relevant factors were not the same ([8, 35]). In our study, the physicians were blinded and presented with identical scenarios differing only in gender. Other studies relied on patients' reports of physicians' decision-making style, so we cannot be certain whether physicians varied in response to the patients' characteristics or whether patients differed in their reporting of physician style (). We too used patients to report physicians' behavior; however, our standardized patients were extensively trained, periodically monitored, and not informed of this study's purpose. Also, our statistical analysis accounted for the clustered data, which should minimize the bias from potential systematic differences in standardized patients' recording of physicians' behavior. Furthermore, standardized patient assessments have been found to be both reliable and valid ().
One possible explanation for the influence of patient gender on physicians' interpersonal behavior includes physicians' conscious beliefs or unconscious stereotypes about the patient. Physicians have been shown to exhibit stereotyping toward patients on the basis of their gender (). Physicians may make assumptions about a patient's level of social support or whether they live alone, and caregiver responsibilities. Also, because physicians ask women fewer questions, they may rely on their assumptions about women (). There is extensive research that all humans share the cognitive strategy of automatically applying beliefs and expectations about groups of people and generalizing these beliefs to all of the individuals belonging to that group ([40, 41]). This strategy is thought to simplify the vast amounts of information to which we are exposed, making our cognitive processing more efficient (). That physicians are immune to social categorization and stereotyping is unrealistic.
An alternate explanation is that patients' communicative behaviors influenced physicians' interpersonal behavior, which may or may not in turn have been influenced by the physicians' behavior ([7, 43, 44]). Four ways by which a patient can influence physicians' behavior include providing a health narrative, asking questions, expressing concerns, and being assertive ([44, 45]). However, our male and female standardized patients presented identical scenarios with the same chief symptom as their standard opening sentence to all physicians, they were trained not to ask questions (other than parroting the physician to help the encounter seem natural) or be assertive, and the only concern that was expressed was via a prompt, “Do you think I need a new knee?” at the end of the visit only if the physician had not recommended and/or freely offered TKA or referral to an orthopedic surgeon. Despite this, their manner of presentation may have been different by virtue of them being a man and a woman. However, had our female standardized patient asked more questions, as evidence shows women tend to do ([8, 35]), this should have resulted in physicians providing more information and having more participatory clinical encounters with her, not less.
A third and final explanation is that physicians are simply not aware of the elements that are important to IDM and lack the interpersonal skills necessary to be able to communicate effectively. Our findings are consistent with other studies that used the IDM scale ([18, 20, 46]) in that IDM for a complex decision was often incomplete. Our IDM scores for the male patient were similar for most elements, but were somewhat higher for discussion of the patient's role in decision making, discussion of the alternatives, and assessment of the patient's understanding of the decision, which may be explained by differences in the studies, specifically that in prior research, the majority of patients were women and most visits were not their first visit to the physician. We demonstrated that deficits in patient–physician communication were significantly greater for the woman. Many advocate communication skills programs for medical students, residents, and practicing physicians that focus on patient-centeredness ([10, 13, 18, 19, 26]) and have even used standardized patients for teaching and evaluating these skills ([13, 19]).
Based on our data, programs such as these will benefit patients in general and women in particular. When the woman was consulted for TKA, compared with the man, physicians were less likely to discuss the nature of the decision, and seldom discussed her role in decision making, explored whether she understood the decision, or elicited her preferences. In essence, since shared decision making takes at least “two to tango” (), most physicians did not ask the woman “to dance.” Physicians also rarely discussed the recovery postsurgery, which we know to be a specific concern for women (). Physicians able to meet patients' informational needs not only satisfy patients' desire for information, but are interpreted by patients to be more concerned about them as people (). Women are more likely than men to prefer an active role in clinical decision making () and a collaborative style of communication () with their physicians. The physicians' failure to include elements or discuss items important to patients may result in them being less likely to accept physicians' recommendations (), and therefore could contribute to the gender disparity in TKA utilization. Patient preferences did not play a role in this study because standardized patients were scripted as being willing to undergo surgery.
A patient-centered approach may be part of the remedy to address the physicians' contribution to the gender disparity in TKA utilization. However, regardless of physicians' recommendation regarding TKA, patient gender influenced physicians' interpersonal behavior. Physicians recommending TKA still included 1.6 fewer IDM elements on average (P = 0.001) and had poorer interpersonal skills (P < 0.001) when the patient was a woman compared to a man. For a complex decision such as arthroplasty surgery or referral to an orthopedic surgeon, all 7 IDM elements should be present for the consultation to be considered complete (). Only for the man did the mean number of IDM elements meet the minimum of 5 for the consultation to be considered adequate for a patient to be able to weigh their increased mobility and decreased pain against the risks of morbidity and mortality of the procedure (). We also found that physicians who include more IDM elements during the consultation are more likely to recommend TKA, suggesting that IDM may be a potential confounder in the model used to assess the effect of patient gender on physicians' recommendations for TKA (). However, regardless of the extent of IDM present in the consultation, physicians remained less likely to recommend TKA to a female patient than to a male patient (OR 2.3 [95% CI 1.2, 4.3], P = 0.03). Therefore, an educational intervention designed to improve IDM and interpersonal skills may not completely address physicians' gender biases and the gender disparity in total joint arthroplasty. In addition to including gender sensitivity in medical curricula (), perhaps what is needed is an additional skills component to patient-based care that provides physicians with the capacity to consciously replace automatically activated stereotypes ([40, 52, 53]).
Contrary to other research findings ([8, 54]), we found that physicians spent less time with the woman than the man during the consultation. IDM increased as the duration of the visit increased. Indeed, for those physicians with mean IDM scores between 0 and 3 compared with physicians with mean IDM scores between 4 and 7, the length of the visit was, on average, 15 minutes and 22 minutes, respectively (P < 0.001). Based on audiotapes of actual orthopedic practice, Braddock et al () also found a greater extent of IDM with an increased duration of the visit. The Medical Outcomes Study also reported the same relationship and proposed that visits of at least 20 minutes may be needed to involve patients effectively in treatment decisions (). Changes in the way physicians are remunerated may alleviate the time pressures and demands of medical practice, thereby reducing cognitive resources and discouraging stereotyping ().
Our study has several potential limitations. First, standardized patients visited only those physicians who agreed to be visited. Our volunteer participants are probably more likely to have a greater interest in improving medical care and may have more skill in assessing and treating patients. If bias were found in these physicians, then probably similar or greater bias would be found in the wider community. Second, we relied on patients to report physicians' interpersonal behavior. However, our standardized patients underwent extensive training and monitoring. Third, since this was a cross-sectional study, we did not have the benefit of observing the patient–physician relationship develop over time. However, even at the first visit, we maintain that at minimum the physician should have included the requisite elements for a basic clinical decision: a discussion of the clinical issue or nature of the decision and elicitation of the patients' preference. Fourth, the study was performed in a single province in Canada. However, because universal health care reduces access barriers to the procedure, Ontario was an excellent setting for this study. Furthermore, because disparities in access to health care have also been found in the US, our results are likely not specific to physicians in Ontario alone. Finally, this was a study of a single surgical procedure. We chose this procedure because of the known gender disparity () and have no reason to believe the results would differ for other procedures or treatment.
In conclusion, physicians included fewer IDM elements and had poorer interpersonal behavior during the consultation when the patient was a woman compared to a man. Although physicians recommending TKA were more participatory, they still included significantly fewer IDM elements when the patient was a woman compared to a man. Also, regardless of physicians' participatory decision-making style and communication skills, physicians were less likely to recommend TKA to a female patient than to a male patient. Our findings suggest that in addition to directly influencing clinical decision making, gender bias may also influence physicians' interpersonal behavior. Because educational interventions designed to improve patient–physician communication are not likely to target physicians' unconscious stereotypes about patients, they are probably not sufficient to address the physicians' contribution to treatment disparities.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Borkhoff had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Borkhoff, Hawker, Kreder, Glazier, Mahomed, Wright.
Acquisition of data. Borkhoff.
Analysis and interpretation of data. Borkhoff, Hawker, Wright.
The authors would like to thank the physicians who participated in this study. We also acknowledge the invaluable contributions made by all members of the “Operation Knee” research team, especially Dorothy Aungier, Marylyn Peringer, Murray Nisker, Len Berk, Mindy Green, Lois MacKenzie, Werner Thom, Gaby Thom, Sam Osak, and Harold Weston, for their work as standardized patients or mock family members. We are also especially grateful to Jennifer Ionson for her work in recruiting participants and data entry, and her unwavering commitment to this project.