Last October, on an airline flight from Detroit to Houston, a passenger became seriously ill, eventually losing consciousness. The attendants asked for medical help. A doctor on the flight came forward to provide assistance. One of the flight attendants refused to allow the doctor to attend the patient without some proof of their qualifications, despite verbal reassurances and the patient’s obvious need. While this was going on, a second doctor appeared who was welcomed and permitted to examine the patient. No qualifications were either requested or offered from the second doctor.

The first doctor was a black woman. The second was a white man.

The attendant apparently had difficulty characterizing a black woman as a doctor, but had no such difficulty with the white man. As you might imagine, this has led to considerable media attention and some rather creative groveling on the part of the airline involved which has vowed to extend its diversity and inclusion training (previously provided only for corporate leaders) to frontline employees.

We’re left nonetheless to ponder how such a thing could happen. The flight attendant involved, we must assume, is neither a raving lunatic nor card-carrying bigot. She is likely a regular citizen who, in a highly stressful situation that required her to make a critical judgment, did so instinctively. In doing so, she exhibited (or made explicit) what could be termed an Implicit Bias.

The notion of Implicit Bias is rather unsettling, particularly to those who quite sincerely believe themselves to be accepting of diversity and inclusion. The concept is that we all harbour prejudicial impressions and attitudes of which we are blissfully unaware. These attitudes, apparently related to personal associations and memories, reside deep in our subconscious but are capable of influencing our decisions and actions without our intention or awareness. Fundamentally, even if we truly, even fervently believe in the principles of equality and attempt to conduct ourselves accordingly, we are all “hard-wired” to identify with and therefore feel greater affinity with those who are like us, and less so with those who differ from us in some way. It can be regarded as a developmental survival adaptation that allowed our ancient predecessors to recognize threats and react quickly to avoid them. It’s what alerts the gazelles to instantly run at the first sign of a lion without taking the time to process the decision. In terms consistent with Daniel Kahneman’s Thinking Slow and Fast approach, it’s the ultimate triumph of Type 1 over Type 2 thinking.

The concept of Implicit Bias is certainly gaining attention and being taken seriously by the scientific community. In a recent edition of Science (352:6289,1035) editor-in-chief Marcia McNutt reports on a forum of editors and publishers of prominent journals convened by the American Association for the Advancement of Science to discuss how Implicit Bias might be countered in the peer review process. Blinding reviewers as to authorship is apparently insufficient.

The Ontario Human Rights Code has, as one if its core principles, the primacy of the consequence or effect of an action over the intention that led to it, surely an acceptance of the influence of subconscious or implicit biases.

In the medical world, there have been a number of rather disturbing reports on the subject. In a study on the diagnostic approach to patients presenting with chest pain (Schulman et al, NEJM 1999;340:618) the authors prepared a series of videos of eight different patients (portrayed by actors) who described their symptoms and medical history. The descriptions and factual information were identical. The authors went to painstaking lengths to ensure the videos were also identical in all aspects, even the facial expressions, hand gestures, background and gowns worn by the patient-actors. The “patients” (pictured below in an illustration taken from the paper) differed only with respect to gender and race.

SchulmanNEJM

The videos were shown to 720 (mostly white) physicians who practiced either Family Medicine or Internal Medicine specialties. The results showed that the physicians were statistically more likely to suspect ischemic disease and therefore order cardiac catheterization in the men than the women, and in the white patients than blacks.

 

A study of 215 surgical attendings and residents at Johns Hopkins (Hader AH et al; JAMA Surg 2015:150:457) used a combination of clinical vignettes and Implicit Association Test (IAT) to assess attitudes and decision making. The instrument identified race and social class biases in most respondents, who were found more likely to suspect alcohol abuse in black patients than whites, less likely to order an MRI in a lower socioeconomic class patient with suspected cervical spine injury, and more likely to suspect pelvic inflammatory disease as a cause of right lower quadrant pain in black than white women.

A similar study was carried out in Oncology programs in Detroit (Penner LA et al; J Clin Oncol 2016;34:2874) involving white oncologists and black patients. Results showed that higher implicit bias in attending physicians (as determined by the IAT survey instrument) was associated with patient interactions that were shorter and perceived to be both less supportive and less effective. Higher Implicit Bias scores were also associated with lack of patient confidence with treatment plans and perceived difficulty in completing the course of treatment.

A recent systematic review of 15 studies of health care providers using the IAT showed low to moderate levels of bias (similar to the general population) in all but one (Hall WJ et al: American Journal of Public Health 2015;105:e60-76).

Many of these studies have evoked considerable criticism. There are certainly counter arguments to be made. Survey instruments like the IAT, even if previously validated, have limitations. The diagnostic process we utilize and teach incorporates known risk factors which are known to vary among genders and racial groups, so what’s perceived to be bias, some argue, may simply be the appropriate application of epidemiological data and “real life” knowledge.

However the real and deeply disturbing concern is that the concept of individual physician judgment or intuition that the profession has valued so greatly, and both clinicians and their patients rely upon to develop effective and efficient treatment decisions, is under threat. If we’re all subject to Implicit Bias, are we capable of making valid decisions on any issue that strays from a strict guideline algorithm? Indeed, will this threaten the confidence clinicians require to make critical decisions?

Fortunately, hope springs forward in the form of several perceptive approaches that are being advanced to counter the threat of Implicit Bias. For those interested in reading further, I will list some relevant papers below, including one by Dr. Geoff Norman and colleagues at McMaster that was published just this month in Academic Medicine. Those articles describe educational, administrative or societal approaches to the issue. What seems to be missing, however, is practical advice to individual physicians and learners as to how they might approach these concerns personally. Taking the liberty to provide a personal and non-expert perspective on this, I would offer the following seat-of-my-pants prescription to combat Intrinsic Bias:

  1. Brutal honesty. It would appear from the literature that no one is immune from these influences. We are all complicated, unique individuals with our own mix of life experiences. Recognizing that these biases exist, are natural and not indicative of a disordered personality of some type, but are nonetheless capable of influencing our decision-making would seem to be the best first step.
  1. Self-awareness. A truly honest exploration of our comfort in engaging patients from various backgrounds and with personal choices or perspectives that differ from our own would seem to be a key step. There’s nothing inherently wrong with not feeling equally comfortable with everyone we encounter. Something will be very wrong if that discomfort influences the care we provide.
  1. Increasing personal diversity experiences. Personal, one-on-one experiences with people from different backgrounds is the most effective way to increase understanding, promote comfort in engaging diversity and break down the threat of intrinsic bias. We need to seek such opportunities for ourselves and, as a school, develop and promote such opportunities for our students. Our Diversity Committee, under the direction of Dr. Mala Joneja, has this goal firmly in their sights and is making considerable progress.
  1. Do the mental “double-check”. When making a clinical decision involving a patient from a diverse group or background, a helpful and simple exercise might be to imagine the same scenario being played out in a patient of your age, gender, culture and socioeconomic/social circumstances. If the decision you would make is different in anyway, there should be a valid medical reason for that different approach. If not, a second thought might be in order.

In summary, Intrinsic Bias is a distinctly uncomfortable topic, particularly for physicians. It is threatening, unsettling, humbling and personal disturbing. But it’s also probably real, and worthy of study and personal reflection on the part of both practitioners and learners.

 

Anthony J. Sanfilippo, MD, FRCP(C)

Associate Dean,

Undergraduate Medical Education

 

Approaches to Implicit Bias:

Byrne A, Tanesini A. Instilling new habits: addressing implicit bias in healthcare professionals. Adv. In Heath Sci Educ 2015; 20: 1255.

Norman GR et al. The causes of errors in clinical reasoning: Cognitive biases, knowledge deficits, and dual process thinking. Academic Medicine 2017; 92: 23.

Penner LA et al. Reducing racial heath care disparities: A social psychological analysis. Policy insights from the behavioral and brain sciences. 2014; 1: 204.

Stone J, Moskowitz GB. Non-conscious bias in medical decision making: what can be done to reduce it. Medical Education 2011; 45: 768.

 

Many thanks to Sarah Wickett, Health Informatics Librarian, Bracken Library, for her valuable assistance in the compilation of information for this article.