The effect of patients' sex on physicians' recommendations for total knee arthroplasty

Methods

Standardized patient scenarios

Both standardized patients memorized identical scenarios that included functional capacity, pain severity, amount of sleep disturbance, use of pain medications and all other treatment modalities (Box 1). Their socioeconomic status was scripted as middle class, and both had physically undemanding part-time jobs. Chronic knee pain was given as their chief complaint so that the physician encounter would be focused exclusively on their osteoarthritis. The scenario was scripted such that usual nonoperative treatment options had been exhausted. To ensure standardization of presentation of the scenario, each standardized patient visited each physician only once.16 We instructed the standardized patients to provide their chief complaint as their standard opening sentence to all physicians. They were given detailed instructions on what information could be shared spontaneously with the physicians and what information was to be provided only when asked. If by the end of the visit the physician had not recommended or offered total knee arthroplasty or referral to an orthopedic surgeon, the patients were instructed to prompt the physician by asking “do you think I need a new knee?”

• Download figure
• Open in new tab
• Download powerpoint

Box 1.

Results

Of the 71 physicians visited by the standardized patients, 67 gave consent to include their data in our study (Table 1). Personal and practice characteristics are representative of practising orthopedic surgeons and family physicians in Ontario.12^,13

As shown in Table 2, more than 90% of orthopedic surgeons considered at least 1 of our standardized patients with moderate knee osteoarthritis to be an appropriate candidate for surgery, compared with 60% of family physicians. Overall, 67% of physicians recommended total knee arthroplasty to the male patient compared with 33% who recommended it to the female patient. Focusing on the discordant pairs for all physicians, 42% of physicians recommended total knee arthroplasty to the male but not the female patient, and 8% recommended total knee arthroplasty to the female but not the male patient.

The overall odds that total knee arthroplasty was recommended to a male patient was 4 times the odds for a female patient (OR 4.2, 95% CI 2.4–7.3, p < 0.001). When these results are expressed as a RR, a male patient was twice as likely as a female patient to receive a recommendation for total knee arthroplasty (RR 2.0, 95% CI 1.5–2.8, p < 0.001). The odds of a male patient receiving a recommendation for total knee arthroplasty from an orthopedic surgeon or being referred for total knee arthroplasty by a family physician were higher than for a female patient (Table 3).

Because few of the participating physicians were female, we could not examine the effect of the sex of the physician on our findings. Of the 12 female physicians, 3 recommended total knee arthroplasty to both the male and female patient, 5 recommended it to the male but not the female patient, 2 recommended it to the female but not the male patient and 2 recommended it to neither patient. Female and male physicians had similar rates of recommendation for total knee arthroplasty; 67% of female and 67% of male physicians recommended total knee arthroplasty to the male patient (p = 0.97), and 42% of female and 31% of male physicians recommended total knee arthroplasty to the female patient (p = 0.47).

Of the 67 physicians, 1 family physician detected the male standardized patient and 1 detected both the male and the female standardized patients, giving a detection rate of 2% (3/134 visits). In total, 1 false-positive detection was reported. The detection rate in our study was much lower than the rates previously reported by other researchers using unannounced standardized patients (0%–18%).16^,18 The primary reason given by both family physicians who detected the standardized patients was the absence of a radiologist's report with the radiographs. Eliminating the data from the 3 visits involving the detection of a standardized patient did not change our results.

Interpretation

Physicians are trained to ensure that clinical decisions are tailored to the needs of the individual patient and reflect the current state of medical knowledge. On surveys, referring physicians and orthopedic surgeons report that a patient's sex does not affect their decision to refer a patient for, or perform, total knee arthroplasty.12^,13 However, we found that, in actual clinical practice, the sex of the presenting patient affected physicians' treatment recommendations. Previous studies that documented sex-related differences in the treatment of osteoarthritis7^,19^,20 were unable to determine the sources of these differences. We addressed this question by using standardized patients with identical clinical presentations of moderate knee osteoarthritis with chronic knee pain who differed only by sex. Our findings suggest that physicians may be at least partially responsible for the sex-based disparity in the rates of use of total joint arthroplasty.7

There are 3 possible explanations for our findings. First, the participating physicians' decisions to recommend total knee arthroplasty may have been based on conscious attitudes or overt discrimination based on sex. Some physicians have been shown to take women's symptoms less seriously and attribute their symptoms to emotional rather than physical causes21 and to refer women less often22 than men for specialty care even when women have a relatively greater degree of disability. A second possible explanation is that the participating physicians' treatment recommendations were a result of an unconscious bias based on gender. Unconscious bias occurs when a patient automatically activates a stereotype in the physician's memory.23 Discriminatory actions stemming from unconscious biases are not deliberate, and physicians would be unaware of them. Our study suggests that physicians are susceptible to the same unintentional gender biases that are pervasive in the rest of society.24 In our study, physicians may not have recommended total knee arthroplasty to the female patient because an unconscious bias resulting from years of experience tells them, or they've heard from other physicians, that women don't receive the same benefit from total knee arthroplasty as men. This inappropriate preconception may be because women typically receive surgery at a more advanced stage of disease than men19^,20 and those with more advanced osteoarthritis have worse surgical outcomes.20^,25^,26 A third explanation is that despite identical clinical scenarios, the presentation style of our male and female standardized patients may have differed because of their gender. Women typically present their symptoms using a narrative style, speaking more openly and personally about their complaints, whereas men typically present their symptoms using a business-like style, describing their complaints in a more factual or reserved manner.27^,28

The results of our study suggest a need for simple tests to detect biases among physicians. Customized implicit association tests29^,30 (similar to those at www.projectimplicit.net) may provide a possible mechanism to identify an unconscious or implicit gender bias. Different strategies at multiple levels may be required to address the sex-based disparity in the rates of use of total knee arthroplasty. At the societal level, strategies may need to address fundamental social and economic inequalities.9 At the health policy level, collecting disparity indicator data31 and increasing the diversity among health care providers32 may help to equalize treatment between the sexes. At the organizational level, gender sensitivity in medical curricula has been introduced.33 Our research findings support the need for intervention strategies directed at the level of health care delivery, through clinician education programs to better inform physicians of the true risks of total joint arthroplasty, when and for whom to consider surgery as well as the potential benefits of early treatment.20^,25^,26 Additional strategies include presenting physicians with data showing that, assuming similar preoperative disease severity, women and men derive similar benefits from total joint arthroplasty.19 However, these intervention strategies may not address the physicians' contribution to the sex-based disparity if the natural social cognition processes that contribute to gender bias are not addressed. Research suggests that individuals have the capacity to overcome their biases and consciously replace automatically activated stereotypes if they are made aware of their biases and are motivated because of personal values or feelings of guilt.23^,34 Thus, we propose developing an intervention that focuses on increasing physicians' acceptance and awareness of the automatic nature of social categorization and stereotyping, and that helps physicians self-recognize and challenge their own conscious and unconscious biases that may be influencing their clinical decision-making. Simultaneous interventions directed at patients to address their contribution to the sex-based disparity are integral to the success of these proposed interventions.

An additional observation in our study was that more than 90% of orthopedic surgeons, compared with 60% of family physicians, considered at least 1 of our standardized patients with moderate knee osteoarthritis to be an appropriate candidate for surgery. This suggests that the recommendation of total knee arthroplasty is the right decision and represents the best care for our standardized patients. Because our standardized patients presented without a radiology report, this may have led to fewer specialist referrals by family physicians who may have been uncomfortable interpreting the radiographs. However, it is more likely that while total knee arthroplasty is generally acknowledged by family physicians and orthopedic surgeons to be appropriate for patients with moderate knee osteoarthritis, family physicians tend to overestimate the risks and underestimate the benefits of arthroplasty.12^,13

There are several potential limitations to our study. First, only 2 standardized patients were included. However, we took several steps to minimize all clinically relevant and extraneous differences between the 2 patients. We included 1 man and 1 woman with moderate knee osteoarthritis and whose cases were confirmed as being identical in terms of disease severity. Both patients received extensive training to present identical standardized scenarios, and their presentation accuracy was formally tested at the beginning of the study and twice during the study by surreptitiously videotaping each patient's performance during office visits with volunteer test physicians. For all patient presentations, the degree of conformity was high. Although the patients' presentations may have been subtly different, it is unlikely that they varied to an extent that would explain the observed differences in the rates of recommendations. Second, the standardized patients visited only physicians who agreed to be visited. However, we do not believe that the physicians who volunteered would be more biased than those who did not. Third, it is possible that some physicians may more fully investigate osteoarthritic symptoms and discuss treatment options during a patient's second visit. However, we believe that this would have occurred infrequently. Furthermore, the standardized patients' script was written such that at the end of the examination, the patients asked if they needed a new knee joint. Fourth, although our sample size was large enough to detect an overall main effect and to perform 1 subgroup analysis, it was not large enough to address all of the potential subpopulations of interest, such as physician age, sex or site of training. Fifth, the study was performed in a single Canadian province. However, because universal health care reduces access barriers to total knee arthroplasty, Ontario was an excellent setting for this study. Disparities in access to health care have also been reported in the United States; thus, our results are likely not specific to Ontario physicians. Finally, this was a study of 1 surgical procedure. We chose to study recommendations for total knee arthroplasty because of the known sex-based disparity in the rate of use for this procedure.7 We have no reason to believe that the results would have been different for other procedures or treatments. Additionally, although the magnitude of the effect differed, we found a gender bias among both referring family physicians and orthopedic surgeons.

In conclusion, physicians were more likely to recommend total knee arthroplasty to a male rather than to a female patient, suggesting that gender bias may contribute to the sex-based disparity in the rate of use of total knee arthroplasty. Our findings suggest that physicians are prone to the same automatic, unconscious and ubiquitous social stereotyping that affect all of our behaviour. Acknowledging that a gender bias may affect physicians' decision-making is the first step toward ensuring that women receive complete and equal access to total joint arthroplasty. Further research is needed to develop interventions to address the physician's contribution to treatment disparities and to develop tools to measure the effectiveness of these interventions.

@ See related article page 723