Defensive Responses to Implicit Association Tests and Bias Awareness in an Implicit Bias Mitigation Training

Introduction

Health and healthcare disparities exist within a system of structural racism that advantages some and disadvantages others. A multilevel approach must be used to remediate health inequities, from the policy and institutional level to individual level interventions.^1,2 Evidence suggests that healthcare clinicians hold negative implicit and explicit biases toward marginalized groups and other disadvantaged populations,^3–8 which can affect clinical decision-making,^5,9 interpersonal communication and patient care.¹⁰ Implicit biases are “attitudes or stereotypes that affect our understanding, decision-making, and behavior, without our even realizing it.”¹¹ In contrast, explicit biases reflect attitudes and beliefs that an individual holds which are available for introspection and which they can report. Research shows that implicit and explicit attitudes may coexist, even among egalitarian individuals.¹⁰ Research shows that becoming aware of personal bias is an important first step toward mitigating the impact of bias in healthcare^12,13 and can result in being more receptive to bias feedback.¹⁴

Given the widespread social structural pathways to inequities, individuals may feel powerless to make a difference. However, increasing awareness of bias, coupled with communication skills training is a promising individual level strategy for narrowing healthcare disparity gaps for racially, ethnically, and socio-economically disadvantaged persons.¹⁵ Implicit bias education for clinicians often focuses on a bias recognition and management model to mitigate the impact of clinician bias on healthcare delivery.¹⁶ Implicit bias education for clinicians can increase awareness of bias,¹⁷ and represents one component of a multilevel approach to address healthcare inequities.¹

The Implicit Association Test (IAT)¹⁸ with feedback on results is often used as a tool in health professions education to stimulate reflection on personal biases.¹⁶ Personal bias awareness can result in greater receptivity to feedback and recognizing acts of subtle bias as acts of racial discrimination.¹⁴ However, the use of the IAT needs to be done with care and an understanding of how learners may react to learning about their biases.¹⁹ Howell et al, 2017 found that when people believe they are less biased than others, a common phenomenon, they may mount a defensive response when confronted with evidence of their implicit bias.²⁰ Greater understanding about how healthcare clinicians react to implicit bias education that includes the IAT with personal feedback is needed.

The purpose of this study was to explore the “I am better-than-average” (better-than-average) effect described in detail by Howell et al, 2017, and defensiveness to IAT feedback²⁰ among residents in internal medicine (IM) and family medicine (FM) and Doctor of Nursing Practice (DNP) students (hereafter referred to as ‘clinical learners’) engaged in a multi-component implicit bias recognition and mitigation educational program, COmmuNity-engaged SimULation Training for Blood Pressure Control (CONSULT),²¹ as part of their required curriculum. In this report we examine the prevalence of defensive response to IAT feedback, personal characteristics associated with defensiveness, perceptions of being better than other clinicians (less bias), and awareness of bias within self, healthcare, and society among clinical learners. We also examine the relationship between better-than-average, defensiveness and IAT scores, as well as the relationship between defensiveness and awareness of bias within self, society and healthcare. Based on our prior research, we hypothesize that the better-than average effect and defensive responses to IAT feedback exists in among clinical residents.

Methods

Study Design

We conducted a cross-sectional study of self-reported implicit bias, explicit bias, and defensiveness to IAT feedback within the context of the CONSULT cluster randomized trial. (ClinicalTrials.gov: NCT 03375918). Details of the CONSULT educational program are described elsewhere.²¹ Briefly, the intervention’s core elements were: a) National Institute of Minority Health and Health Disparities-funded e-learning modules to build knowledge about health disparities, implicit bias, and patient-centered communication skills;²² b) IAT with feedback to raise learners’ personal bias awareness;¹⁸ and c) bias-mitigating communication skills training with practice in clinical simulations with standardized patients who were drawn from the local community.²³

Study Setting and Data Collection

Study Setting

CONSULT was integrated into the IM and FM residency programs and DNP curriculum at a medical and nursing school affiliated with a disproportionate share hospital. The patient population of the medical center included ~27% non-White and 21% Hispanic/Latinx persons; ~34% of the medical center’s surrounding county’s residents do not speak English at home and 43% were low income.

Eligibility Criteria

Training programs assigned their own learners to participate in CONSULT. Participants provided informed consent for use of their data for outcomes analysis.

Measures

Implicit Association Test (IAT)

We asked participants to complete two race and ethnicity IATs [Black/White, Hispanic-Latinx/White]¹⁸ and two Race and Medical Compliance IATs [Black/White, Hispanic-Latinx/White].²⁴ Corresponding explicit bias questions (described below) were presented as per the standard protocol of Project Implicit, the IAT vendor. The Race Attitude IAT¹⁸ measures implicit attitudes about race. The Race and Medical Compliance IAT²⁴ measures implicit beliefs about race and ethnicity and patient medical compliance. The IAT asks participants to pair facial images or representative words with value categories of good versus bad, positive versus negative.

Defensiveness

We asked three questions from Howell & Ratliff, 2017,²⁰ about participants reaction to the IAT, assessed on a 4-point scale (Strongly Disagree, Disagree, Agree, Strongly Agree) to measure degree of defensiveness after taking the IAT. The defensiveness questions are: “The IAT does not reflect anything about my thoughts or feelings unconscious or otherwise”, “Whether I like my IAT score or not, it captures something important about me” and “The IAT reflects something about my automatic thoughts and feelings concerning this topic.”²⁰

Explicit Bias Measures

As per the standard protocol of Project Implicit when administering the IAT, participants were also asked about their own explicit beliefs.²⁵ We collected the following explicit measures to fully characterize our sample’s attitudes and beliefs. Participants indicated their personal explicit attitude on a scale ranging from 1 = I Strongly Prefer Black [Hispanic-Latinx] People to White People to 7 = I Strongly [prefer White People to Black People] with a mid-point of “I like White People and Black [Hispanic-Latinx] People equally”. We asked explicit race and ethnicity (Black [Hispanic-Latinx]) medical compliance questions, “In general, members of which group are more likely to be regarded as compliant patients?”

Better-Than-Average

In addition, we asked the same explicit questions that corresponded with the content of each IAT for “other health care providers” and calculated a difference score [Others minus Self] for better-than-average.^20,26 A positive score greater than zero demonstrates participant’s perception that others have more bias than they do, or that they are ‘better-than-others’.

Bias Awareness Scale

Trainees were asked to complete a 7-item Bias Awareness Scale, with items assessed on a 6-point scale (strongly agree to strongly disagree); higher scores indicate high levels of bias awareness in the domains of society, healthcare, and within oneself.²⁷

Data Analysis

We used descriptive statistics (means, standard deviations, and frequencies) to characterize the clinician sample, mean scores for defensiveness, and item responses for the bias awareness scale. The IAT D effect was calculated using the standard IAT scoring algorithm.²⁸ The IAT D score ranges from −2 to +2, with zero indicating no relative preference between White people and Black [Hispanic or Latinx] people. An IAT D score that differs positively from zero some degree of implicit preference for White people. Cohen’s d, a standardized effect size measure, was calculated for each of the implicit and explicit measures. Cohen’s d is interpreted as d of 0.20 = small effect, d of 0.5 = medium effect, and d of 0.80 = large effect.²⁹

For the explicit measure we coded the seven-point response scale to range from −3 to +3, with zero indicating no relative preference for White people over Black [Hispanic-Latinx] people. An explicit measure mean that differs positively from zero indicates an explicit preference for White people over Black [Hispanic-Latinx] people. We compared means for the implicit and explicit measures for the complete sample of participants and by defensive/not defensive. We calculated the defensiveness score according to Howell, 2017 by averaging the three items.²⁶ Higher scores signified greater defensiveness.

We computed a “better-than-average” index from the implicit bias and explicit bias measures. Specifically, we subtracted people’s pro-White estimates for themselves from their pro-White estimates for ‘other health professionals’.²⁰ To the extent that people indicated that they had less anti-Black [Hispanic-Latinx] bias than did ‘others’, we considered them to be demonstrating the better-than-average effect. We calculated the Wilcoxon Rank Sum for each comparison of implicit and explicit measures between self and others to identify whether the difference was statistically significant.

We then examined the association between defensiveness and IAT scores, as well as between defensiveness and measures of bias awareness [personal, societal and in healthcare], using t-tests. The study was reviewed and approved by the UMass Chan Medical School Institutional Review Board (IRB). This study complies with the Declaration of Helsinki.

Results

By September 1, 2019, 125 trainees were assigned to the CONSULT training, and 61 enrolled and completed all assessments.

Defensiveness

We found that 18.0% of participants had a defensive response to feedback on their IAT scores. (Table 1) Characteristics of participants by defensiveness is shown in Table 2. There was no significant difference in participant characteristics by defensiveness except for age; those who were not defensive were younger than those who were defensive (34.7 years versus 29.8 years, p= 0.004). There were no Hispanic-Latinx people in the defensive group compared to 12% of the not defensive group were Hispanic-Latinx people (p= 0.046).

Table 1 Feedback to the IAT Defensiveness Measure (N=61)

Table 2 Characteristics of Trainee Participants (N=61)

Implicit Attitudes

We found weak to moderate implicit attitudes favoring White people on the Black/White and Hispanic-Latinx/White race attitude IATs and beliefs about race and medical compliance endorsing White people as more medically compliant (Cohen’s d range: d=0.37– d= 0.63 respectively). (Table 3) We found no difference in any implicit bias measures by defensive and not-defensive to feedback on IAT scores.

Table 3 Implicit and Explicit Measures and Defensiveness

Explicit Attitudes

Participants reported little explicit preference for Black versus White people, and Hispanic-Latinx versus White people (Cohen’s d=0.22 and d=0.20) respectively. (Table 3) They ascribed strong explicit race and ethnicity bias to “other health professionals” (Cohen’s d=1.11 and d=1.24) respectively. On the medical compliance IATs, participants reported that both they and other health professionals hold a strong association of White people as more compliant than Black and Hispanic-Latinx people, with Cohen’s ds ranging from d= 0.78 to d=1.24. Those who were defensive held a moderate explicit association of White patients as more medically compliant than Hispanic-Latinx patients, compared to strong bias expressed among those who were not defensive (Cohen’s d=0.50, and d= 0.91, p= 0.048). (Table 3) The same pattern held true for perceptions of medical compliance among other clinicians (Cohen’s d=0.67, and d= 1.22, p= 0.01). Those participants who were defensive reported that other clinicians held moderate explicit preference for White people and those who were not defensive reported that others held strong explicit race bias (Cohen’s d=1.27, and d= 0.55, p= 0.04).

Better Than Average Beliefs

On all measures of explicit bias of “self” and “other healthcare professionals (others)” participants rated their own attitudes and stereotypes less biased than those of “others.” Participants believe that they are “better-than-average”, regardless of their defensiveness to feedback on their IAT scores.

Figure 1 shows the distributions of explicit self-reports and expectations for others among participants in the study for four assessments: White vs Black, White vs Hispanic-Latinx; Medical Compliance among White vs Black, and Medical Compliance among White vs Hispanic-Latinx.

Figure 1 Better than Average beliefs, by 4 types of IAT assessments. For each IAT assessed, explicit bias is also measured. The better than average measure is based on a comparison of respondents’ self-reported explicit attitudes of bias (blue bars) compared to their perception of explicit attitudes of bias among their peers (red bars). The figure shows better than average beliefs by each of the 4 IATs assessed, including White/Black bias, White/Hispanic-Latinx bias, White/Black compliance bias, and White/Hispanic Latinx compliance bias.

Bias Awareness

On the bias awareness scale,²⁶ participants were aware that there is bias in society and in healthcare. However, participants agreed that they are objective in their decision making, indicating they held very little awareness of possible personal bias. There was no significant difference in any bias awareness category between those who demonstrated defensiveness to feedback and those who did not (Table 4).

Table 4 Defensiveness to Feedback on the IAT and Bias Awareness

Interaction of IAT Scores, Defensiveness and Better-Than-Average

There were no statistically significant interactions, between IAT scores, defensiveness, and better-than-average for Race and Medical Compliance IATs (p=0.9 and p=0.7, respectively for the Black and Hispanic-Latinx conditions) or for Race Attitude IATs (p=0.4 and p=0.5) (Figure 2).

Figure 2 Relationship between Better-than-Average, defensiveness and IAT scores, by 4 types of IATs.Better-than-average Measure: [Bias Other Healthcare Professionals – Bias Self].These figures compare those with high better-than-average to those with no better-than-average. A positive score greater than zero demonstrates participant’s perception that they are less biased or “better” than others.

Discussion

In implicit bias education, a gap exists between what is either empirically or theoretically known to be effective, and what is actually delivered in the course of educational program implementation, particularly in the important but charged domain of implicit bias in healthcare. Awareness of personal bias may be a key step toward addressing the issue.¹³ In the context of CONSULT, we examined whether clinical learners believed that they were less biased than their et.al (better-than-average), their bias awareness, and determined whether they had a defensive response to their IAT feedback. Recognition and acceptance of personal bias can be an emotionally uncomfortable experience for some.¹² If an individual accepts responsibility for holding implicit bias their discomfort can result in behavior and attitude change. If they do not accept personal responsibility for their feedback the result can be defensiveness.¹²

Almost one-fifth of our sample demonstrated defensiveness. Defensive responses to personal implicit bias are a barrier to engaging in implicit bias education and reduction of defensiveness should be attended to in education design. Vitriol & Moskowitz (2021) found that defensiveness to implicit bias feedback can be reduced by education that increases perceptions that bias can be controlled, reducing blame and moral threat.³⁰ If defensiveness does occur, one study found a paradox, that despite discrediting personal feedback on the IAT, medical residents reported that taking the IAT would impact the way they practice.³¹ Thus, although positive reactions to IAT feedback are ideal, it appears that negative reactions to IAT feedback may motivate some individuals to engage in anti-bias education.

Our findings show that clinical learners are aware that bias in society and healthcare exists. Fewer acknowledged that bias can influence their own decision making. This lack of awareness of personal bias underscores the importance of the didactic component of the CONSULT educational program that aimed to increase knowledge and awareness of disparities in healthcare and the complex interplay between implicit bias and clinical decision-making. For the implicit medical compliance and Hispanic-Latinx stereotype, we found that participants who were not defensive reported a stronger explicit belief of White patients rather than Hispanic-Latinx patients being more medically compliant than did defensive participants. We speculate that these non-defensive participants may have been more willing to admit their personal bias than those who were defensive.

To our knowledge, ours is first study to measure the phenomenon of “better-than-average”²⁰ in regard to racial and ethnic attitudes and race/medical compliance beliefs among clinical learners. We found that the majority of participants perceived themselves as less biased than other health professionals. This finding is consistent with prior studies of other individuals who feel that their personal racial bias is not as strong as that of their peers.^20,27 Future research should examine the impact of learners’ better-than-average beliefs and clinician’s response to implicit bias educational interventions, clinicians’ receipt of feedback on personal bias, and patient reported experiences in the health care system.

Taken together, our findings underscore the importance of proactively addressing clinical learner defensiveness to feedback on their biases, incorporating mitigation of defensiveness, and understanding better-than-average beliefs into educational interventions to mitigate bias in clinical encounters. Since this study was originally conducted, and in response to feedback from learners and faculty and in light of growing awareness of the need to address learner defensiveness in implicit bias education,¹⁶ we re-designed the subsequent CONSULT educational program to reduce the number of IATs requested and we made the IAT an optional component.³² This work leads us to question what constitutes best practice for designing an implicit bias recognition and mitigation program that allows for and addresses clinical learner defensiveness?

Emerging evidence from another program called “African American Community Speaks” (AACS), a program that aims to present community member generated lessons about racism and end-of-life communication challenges within a clinician communication training, offers an alternative strategy for raising learners’ awareness of bias while creating space for defensiveness and allowing it to dissipate.³³ AACS has been delivered to more than 189 clinicians from 38 institutions and has an increasing demand for use.³⁴ Some main differences between CONSULT and AACS is that: 1) AACS does not use the IAT as an awareness-building experience, and 2) AACS is designed based on Kolb’s experiential learning theory, which has a holistic perspective that includes experience, perception, cognition and behavior.³⁵ Specifically, AACS uses the first stage of “experience” as an awareness-building experience by asking learners to reflect on a “cringe-worthy” encounter that may be evidence of their potential for biased behavior in care delivery. The AACS trainer then asks learners to hold that “experience”, while hearing external information about the stories of bias from community-members, and then ask the learners to reinterpret their existing experiences in light of new information. Kolb’s learning cycle has four stages: concrete learning, reflective observation, abstract conceptualization and active experimentation. Early AACS findings suggest that defensiveness dissipates during the trainings and often transforms to gratefulness for learning new skills. (Personal communication) Future research needs to identify evidence-based bias defensiveness mitigation strategies.

Our study has several limitations. First, this was completed at a single academic institution, which limits generalizability. Second, this was completed in 2019, before the George Floyd tragedy ignited a groundswell of racism awareness and diversity, equity, and inclusion programs to address its effects. Third, the trial from which these data were collected was powered to examine clinical outcomes for patients of the participating clinicians who are presented in this manuscript. Since many participants did not provide consent to use their data for this analysis, our findings may be underpowered. Finally, our learner participants were racially, ethnically, and linguistically diverse; this sample may not be widely generalizable to other training programs. Mitigating this, our study is the first to examine better-than-average reactions and defensiveness in clinical learners in the contact of IAT use.

Conclusion

While the etiology of healthcare disparities are multifactorial and embedded in systems of inequity, there have been extensive efforts to combat contributing factors such as differential access to care, differences in healthcare outcomes for racial, ethnic and poor populations persist.¹⁵ Working at the individual level, our effort was to implement theoretically sound and novel approaches to building clinician skills to mitigate bias and enhance patient-clinician communication across intercultural differences. The integration of IATs and bias awareness is increasing in healthcare system education, knowledge of how to efficiently and effectively engage learners in all stages of their career in efforts to be aware of and work on mitigating their implicit biases in a psychologically safe and respectful way are still largely unknown. Our paper provides additional areas of attention such as clinicians’ defensiveness to feedback and personal bias awareness when designing implicit bias recognition and mitigation education.