It Is Time to Rethink Our Approach to Bias in Medicine

For most of the twentieth century, anyone attending a symphony orchestra in the United States would have been met by a sea of mostly white, male instrumentalists, each hand-selected from an elite lineage of students trained by master musicians. Starting in the 1970s, however, in an effort to introduce the democratic process to auditions and hire the most gifted musicians, major orchestras began to adopt the use of screens to blind selection committees to the identity of those auditioning. The effect? In a seminal study on bias in hiring practices, researchers discovered that the use of blinded auditions increased the likelihood that a female musician would advance to the next audition round by 50% and the likelihood of being hired after the final round by 30%.1 So, what happened? The researchers suggest that the introduction of screens redirected committee members’ focus to a more objective measure of participants’ qualifications—the sound of the music itself. As a result, the new audition protocols eliminated the influence of any biases that the committee members may have had about the skill of the musicians based on gender. WHAT IS IMPLICIT BIAS? This instance of gender bias is just one example of stereotypes that lead to discrimination. In recent years, the COVID-19 pandemic and the Black Lives Matter movement have prompted important cultural conversations about the pervasiveness of other biases, such as those based on race, ethnicity, and immigration status, among others. Research shows that these biases exist on a spectrum of our awareness. On one end are explicit forms of bias, such as overt racism, sexism, homophobia, or transphobia. The other end of the spectrum, known as “implicit bias,” is more insidious because it operates subconsciously. Our brains must process an endless stream of information. As a result, they tend to make generalizations based on pattern recognition to help us process the world more efficiently. These patterns are influenced by the cultural messages to which we are exposed over the course of our lives in the forms of the media we consume, the people with whom we spend time, and the education we receive. Or, as Harvard Professor of Psychology Mahzarin Banaji, PhD, has described the phenomenon, we all carry a “thumbprint of the culture” in our minds.2 In 1998, Banaji and her colleagues Greenwald and Nosek developed the Implicit Association Test (IAT)3,4 to study the positive and negative values that people subconsciously associate with various personal characteristics such as race, ethnicity, gender, and age. Despite its limitations, the IAT has been the standard for evaluating implicit biases since its development and has been fundamental to the scientific community’s understanding of how bias influences cognitive processes. Research using the IAT has shown that children as young as 6 years of age display racial preferences.5 In addition, people can—and often do—hold explicit egalitarian values while also exhibiting implicit biases against others. This means that people may still unwittingly discriminate against others even as they strive to treat everyone fairly. Although implicit bias operates outside of our conscious thought processes, it can still have a powerful influence on our judgment and behaviors, particularly during stressful situations. Under pressure, the mind resorts to using mental shortcuts—a process that has been defined as “system 1,” or “thinking fast,” by Nobel Prize-winning economist Daniel Kahneman in his book Thinking, Fast and Slow.6 System 1 thinking is automatic, relies on pattern-based generalizations, and occurs outside of our awareness; however, though such mental shortcuts may help us make decisions quickly under stress, they also render us prone to errors in judgment. By contrast, “system 2,” or “slow thinking,” is more deliberate and uses our conscious values. System 2 requires more cognitive effort but is less likely to be influenced by our underlying biases (Fig. 1).Fig. 1.: System 1 versus system 2 thinking. (From Kahneman D. Thinking, Fast and Slow. New York: Farrar, Straus and Giroux; 2011.)Slowing down during situations that normally elicit our automatic thought processes can activate system 2 thinking and help overcome the influence of bias on our judgment. This phenomenon has been best studied in the context of policing, where racial biases play an important role in the disproportionate deaths of unarmed black men shot by the police. Laboratory simulations demonstrate that when officers are fatigued or forced to react quickly, they are more likely to erroneously shoot unarmed black targets. However, when well-rested, officers do not display race-based bias in shooting decisions and are no more likely to shoot unarmed black targets than unarmed white targets, suggesting that fatigue inhibits officers’ cognitive control and increases their reliance on heuristics.7,8 Future strategies for reducing racial bias could include interventions that buy police extra reaction time, reduce fatigue levels, or challenge officers to practice critical reasoning under high cognitive loads.9 Efforts to promote more conscientious thinking have already proven successful. In 2017, the Oakland Police Department added the question “is this stop intelligence led?”—meaning based on evidence rather than intuition—to the forms officers complete when pulling people over to encourage reflection about the motivation behind traffic stops. Over the following year, traffic stops of African Americans decreased by 43%.10,11 Interventions that activate system 2 thinking could similarly mitigate the impact of bias in other industries, such as medicine. BIAS IN MEDICINE Medicine is also a high-stress environment. Physicians and other members of the health care team are just as likely as the general population to exhibit implicit bias12,13 and can unwittingly influence patient care and outcomes as a result. Medicine often requires physicians to make judgment calls in the face of uncertainty based on incomplete information. Furthermore, physicians, like police officers, at times must make life-or-death decisions on little to no sleep. Under these conditions, the mind is at risk of resorting to system 1 thinking and filling in the gaps of missing information with biased assumptions. Examples of bias abound in medicine. Historically, most research participants in medical studies have been white males, creating a data gap about disease presentations and best practices for female and minority patients. Assumptions about how a patient with a particular condition should appear or behave can lead to delays in treatment or misdiagnosis with potentially deadly consequences. For example, research shows that racial and ethnic minorities are less likely to receive pain medication than white patients in the emergency department.14 In laboratory simulations, black patients are also less likely to be prescribed potentially life-saving thrombolytics when presenting with chest pain.15 Women with acute coronary syndrome are also more likely to experience delays in diagnosis because of stereotypes about women’s emotionality and the classification of their common presenting symptoms as “atypical.”16–18 BIAS TOWARD PHYSICIANS AND PHYSICIANS-IN-TRAINING Bias in medicine is not just a problem for patients—it also impacts physicians and trainees. A recent survey of plastic surgery residents revealed that 68.8% had experienced a “microaggression” within the past year.19 Microaggressions are subtle comments or actions that imply devaluation or general disrespect toward someone based on their identity and are driven by the perpetrator’s underlying biases against a marginalized group. Surgical residents who are female, racial, or sexual minorities are more likely to experience microaggressions.19,20 Moreover, trainees who hold more than 1 identity historically underrepresented in medicine—such as black female residents—are even more likely to experience discrimination because of the intersection of their marginalized identities. In a recent survey of surgery residents across the country, microaggressions most commonly came from patients, staff, and faculty, respectively.20 Classically, this might take the form of a patient directing questions to the white male medical student instead of the female or minority attending surgeon. Other examples include comments such as “you don’t look old enough to be a doctor” to a young trainee or questions such as “where are you really from?” to a physician of an ethnic minority background raised in the United States. Over the course of surgical training, odds are you have likely witnessed or personally experienced comments such as these from patients, colleagues, or superiors. Condidered at face value, such comments may seem to be benign or even intended as compliments; however, they can be hurtful, marginalizing, and undermining of a physician’s expertise because they are constant reminders that a person does not fit the expected image of a surgeon. Compounded over time, microaggressions erode physician wellness and are associated with increased rates of burnout.20,21 Bias does not end after medical training. Data show that faculty of color are promoted at lower rates than white faculty and leave academic medicine at higher rates.22 Female attending physicians are also less likely to be introduced by their titles compared with male attending physicians.23 In a study of resident teaching evaluations, researchers identified that female faculty receive significantly lower median teaching scores compared with their male counterparts in male-dominated specialties—the vast majority of which are surgical subspecialties.24 These findings suggest that trainees may penalize female faculty for failing to meet the gender stereotypes associated with male-dominated specialties. The dispersion in scores is troubling because teaching evaluations play an important role in faculty promotions. Bias in teaching evaluations threatens career advancement for female faculty and could be a contributing factor to the relative lack of women in leadership positions within male-dominated specialties. Similar disparities likely exist for faculty of other minority groups. CONFRONTING BIAS IN MEDICINE Ultimately, bias in medicine is harmful to both patient and physician well-being and compromises the quality of medical care. Clinicians who receive high implicit bias scores on the IAT are rated lower on patient-centered care and communication by black patients.25 Because healthy patient-physician relationships translate to better patient outcomes, implicit biases can jeopardize physicians’ ability to provide high-quality, equitable care to all patients. Moreover, delays in diagnosis and treatment because of bias have potentially deleterious consequences. In recent decades, researchers and practitioners have been working to find solutions to counteract implicit bias. The most common strategy has been the use of debiasing workshops that educate participants about implicit biases with the hope that greater awareness will eliminate biased behavior; however, these workshops are highly variable and have no regulatory oversight to ensure that they are effective and evidence based. Most programs also have little to no follow-up on outcomes beyond the immediate postworkshop period. Importantly, debiasing workshops have not been shown to have any meaningful reproducible or durable effect on biases or behaviors over time.26 In line with these limitations, most surgical residents do not think that current debiasing training is helpful.20 In light of these considerations, it is time that the medical community reassesses the way it approaches bias. Although research should continue ongoing efforts to identify existing health disparities, future efforts should prioritize the implementation and systematic evaluation of pragmatic interventions that have an immediate effect on the biased behaviors directly impacting patient care and physicians. Although more research is needed to develop evidence-based protocols, the most effective solutions will likely combine systemic changes with interventions at the level of individual physicians. What follows is a list of action items that physicians can begin incorporating into their daily clinical practice in addition to suggestions for systems-level strategies with the potential to mitigate bias in medicine. PHYSICIAN-LEVEL INTERVENTIONS Take an IAT to Identify Your Underlying Biases It is important to recognize that everyone has biases. Recognizing our personal biases is a first step toward reducing their impact on patient care. Consider taking an IAT to learn more about your biases and identify areas for improvement (available online at https://implicit.harvard.edu/implicit/takeatest.html). Communicate Mindfully Aim to be inclusive with your language. Make a habit of asking and then using a patient or colleague’s preferred name and pronouns (eg, she/her, he/him, they/them). Using the correct pronouns can go a long way in making patients, trainees, and other members of the health care team feel included and valued. Physicians are also at risk of unintentionally perpetuating their biases through documentation in the electronic health record. Research shows that patients who are described using stigmatizing language in medical records receive less aggressive pain management compared with patients described in neutral terms.27 Avoid using phrases that have the potential to be demeaning. For example, “patient with substance use disorder” is preferable to “substance abuser” because it recognizes that addiction is not a person’s identity. Other phrases that should be avoided include anything that might question the patient’s credibility or portray the patient as “difficult.”28 Finally, be mindful of what you say in the presence of impressionable trainees. A study of implicit racial bias among medical students revealed that medical students who overheard attending physicians or residents make negative comments about black patients demonstrated significantly greater implicit racial bias after clinical rotations compared with their first year of medical school.29 This is a solemn reminder that physicians must model the behavior they wish trainees to emulate. Practice Active Listening Clinicians with higher bias scores on the IAT are more likely to verbally dominate patient encounters and engage in less shared decision-making, regardless of a patient’s race.30 Shared decision-making should be prioritized in all clinical encounters because patient-centered communication has been shown to inspire patients to feel higher levels of confidence in their doctor and improve treatment adherence.30–33 Surgeons can ensure they do not dictate patient encounters by engaging in active listening strategies such as those detailed in Figure 2.Fig. 2.: Active listening strategies for patient encounters.Become an Active Bystander Witnessing biased behavior in the workplace can be uncomfortable. When an authority figure speaks up, it can be extremely validating for trainees who might feel they do not have the power to challenge the inappropriate behavior themselves. As leaders of the health care team, attending physicians or senior residents often wish to say something in these moments but may not know how to best respond. For this reason, physicians at the University of Washington have created an in-the-moment protocol to help bystanders combat patient-initiated bias toward trainees.34,35 The protocol emphasizes the importance of addressing the inappropriate comment, refocusing the conversation on the patient’s health, and debriefing after the encounter. The example script includes phrases such as, “let’s keep it professional” or “I am/we are here to focus on your health” (Fig. 3). Consider rehearsing the suggested phrases out loud so they feel more natural and keeping them somewhere they can be easily referenced.Fig. 3.: Adaption of the University of Washington’s bystander intervention for patient-initiated microaggressions. (Adapted from Shankar M, Albert T, Yee N, Overland M. Approaches for residents to address problematic patient behavior: before, during, and after the clinical encounter. J Grad Med Educ. 2019;11:371–374; and Overland MK, Zumsteg JM, Lindo EG, et al. Microaggressions in clinical training and practice. PM R 2019;11:1004–1012.)SYSTEMS-LEVEL INTERVENTIONS Prioritize Diversity within Organizational Structures Institutions should work to incorporate diversity into their organizational structure. Greenwald et al. have proposed the model of the chief diversity officer (CDO) as one possible strategy.26 The authors emphasize that the CDO must be empowered with the resources to analyze disparities within their institution and the support to implement new policies to reform identified areas for improvement so that the CDO position is more than just a figurehead. Increase the Number of Counterstereotypical Leaders and Mentors One of the few strategies shown to reliably reduce implicit bias is exposure to counterstereotypical leaders. In a study of female students in science, technology, engineering, and mathematics, researchers demonstrated that exposure to female professors and mentors reduced internalized biases among female students, increased their sense of belonging, and improved their retention in engineering careers.36 More importantly—and unlike debiasing workshops—these outcomes were sustained years after the mentoring relationships ended.37 In the medical field, teaching faculty and hospital leadership should reflect the diversity of the students and patients they serve. and mentoring programs are potential strategies for of minorities in leadership positions and trainees to leaders with whom they can the to and The problem of bias among selection committees is not to In a researchers to potential but or to the of the The were to be from 50% more than with a that across and faculty selection committees are at risk of similarly implicit biases to color the way they for medical training and faculty positions is highly it is to in the way musicians have been using One potential strategy for a to the process is the use of a strategy for reducing hiring bias by using a of questions and to a system based on objective (Fig. processes, which such as and United States scores and take into and have also been shown to the of by underrepresented interventions for implicit to Clinicians in the Professor Banaji has suggested that within electronic could be to physicians about biases when known to be prescribed at rates to such as pain like this could physicians to use system 2 potentially any underlying biases that might normally influence their clinical Increase within Bias in medicine in because the information that medical students learn is based on of research about the of white, male For example, medical have historically to include color in about even though symptoms may on For the any medical student can you is for on In recent years, have this and efforts to include with a greater of in their teaching from other should out to similarly include patient within their (Fig. Implicit bias is a health that is to patient-physician to physician and health Despite their bias awareness workshops have not been shown to reproducible or Implicit bias action with reduce the to patients and physicians, health care and must prioritize interventions at both the individual and systemic levels that on biased behaviors and patient Finally, it is that these interventions be studied to best practices for meaningful from the of book from and and a research from to study The authors have no relationships or of to The authors the and from of at the of