How I Teach Racial Categorization in Reference Equations Used to Interpret Spirometry

Abstract

Race has been used in medical algorithms, including reference equations for spirometry, following the assumption that race is a biological concept due to the finding that Black American and Asian populations have lower lung function (i.e., forced expiratory volume in 1 second and forced vital capacity) than White populations. Recognizing that race is a social construct and its use in medical algorithms may contribute to health disparities, this practice has been questioned. This inquiry, combined with recent evidence, led the American Thoracic Society to recommend using race-neutral reference equations for spirometry in 2023. In this article, we share our approach to teaching spirometry interpretation, focusing on the factors that determine lung function and the effects of using race-based or race-neutral reference equations. We explicitly define race as a social construct and review the likely reasons for the differences in lung function found between racial groups. We use an interactive and case-based approach to facilitate learning. As part of our interactive approach, we incorporate strategies to enhance psychological safety for the learners. This is particularly important for this topic because discussions of race and health inequities can be contentious and uncomfortable and have the potential to exacerbate the difficulties faced by learners from groups that are underrepresented in medicine. Our approach is intended for a small group of graduate medical learners and could be presented during formal didactic instruction or informally during clinical teaching.

Who Are the Learners?

This approach to teaching the relationship between race and pulmonary function is intended for graduate medical learners such as internal medicine and thoracic surgery residents and for pulmonary fellows. It is also appropriate for trainees and attending physicians in other specialties, who employ pulmonary function tests results in their work.

This group of learners is an ideal audience for rapidly changing material. Some work suggests that recent graduates from medical school have a greater openness to embracing clinical practice guidelines (1). Cohort effects are also observable in social views that appear to parallel the general population (2). These points are relevant because of the gap in how the understanding of race has evolved in many social sciences compared with the life sciences and medicine (3). Graduate medical learners are at an important intersection. Having started medical school 5–8 years earlier, many graduates will have been taught algorithms that portray race as a biological variable reflecting fixed characteristics despite the evidence against this (4–6). Simultaneously, given their age, they may have greater exposure to more current understandings of race and disparities than others in our profession, better equipping them to understand the impact of racial categorization and structural racism in health care (7). Reinforcing this possibility, our content clearly describes race as a social construct that reflects social practices and institutions.

Graduate medical learners can then transfer the knowledge learned here to clinicians who have been practicing longer. These clinicians are even more likely to have been taught to accept racial differences in pulmonary function as normal and to equate race with genetics. Over time, we expect the material to become more appropriate for trainees earlier in their education, replacing erroneously determinist frameworks as medical understandings of race catch up to those in other disciplines.

What Is the Setting?

We envision presenting on a real-time drawing medium (e.g., whiteboard) to a group of approximately 25 learners. The material is adaptable to teaching during clinical service or formal sessions such as residency program conferences. It will take approximately 1 hour to review all the content presented below and in Table 1.

Table 1.

Summary of discussion points and questions for teaching about racial categorization in reference equations used to interpret spirometry

Panel	Concept/Question	Discussion Points
A	Uses of PFTs	PFT interpretations typically categorize the results into one of a few patterns, the most common being normal, obstruction, and restriction. These then help in making diagnoses, formulating treatment plans, and evaluating the underlying cause.
A	Example 1: dyspnea on exertion. What does this spirometry show?	Spirometry is normal. FEV1/FVC ratio, FEV1, and FVC are all above the lower limit of normal (above z-score of −1.645). FVC and FEV1 percent predicted are within 80–120%.
B	Factors used to estimate lung function. What are the demographic data that are used to calculate the predicted values?	We interpret PFT results by comparing the actual values achieved by the patient to the predicted values. The predicted values come from reference populations thought to have normal lung function. To calculate predicted values, we use the patient’s age, sex, and height because these are all associated with lung function. For this example, we used a 177-cm-tall, 50-year-old man. Sometimes race or ethnicity is also used, and this session will focus on understanding why and if race/ethnicity should be used to calculate predicted values. Note: we do not use BMI or weight. These factors do not add to the precision of the predicted values (8), and the reduction in lung function seen in people with very high or very low weight is considered pathological (9).
B	Lung function follows a normal distribution	For people of a given age, sex, and height, lung function will follow a normal distribution, similar to most physiologic variables. We interpret lung function as normal when the values are above a defined cutoff point.
C	What is the cutoff point used to define lung function as normal?	We used to use percent predicted to interpret spirometry, with 80–120% predicted being considered normal. Now, we use z-scores. A z-score is assigned to a given value based on how far, in standard deviations, the value lies from the mean. We typically define 95% of a population to be normal. This results in using a z-score of −1.645 to define the lower limit of normal for PFT values, as 5% of values on the normal distribution lie further from the mean than this (10).
D	Race and ethnicity have been used to estimate lung function	Many studies conducted over more than 100 years have found differences in the average FEV1 and FVC between racial groups. We will focus on GLI 2012 data because this combined many existing data sets to generate one of the largest, and the GLI 2012 equations were the previously recommended equations. The GLI 2012 provided equations for White, North East Asian, South East Asian, and African American groups. The average values for the White group were the highest, those for the African American group were the lowest, and the North and South East Asian groups were in between (12).
D	Race has been used in several clinical algorithms	Race has been used in clinical algorithms for eGFR, cardiovascular risk, and vaginal birth after cesarean section. Race has been removed from these algorithms (14).
E	What factors other than race might explain the finding that Black and Asian people have lower lung function than White people?	A number of factors have been associated with lung function, including air quality/pollution (18, 19), childhood respiratory infections (17), nutrition (22), prenatal exposures (23), socioeconomic status (20, 21). In the United States, many of these factors are experienced differently by people of different races, for example Black, Asian and Hispanic U.S. residents typically experience higher levels of pollution than White residents (24). This situation has roots in discriminatory policies such as redlining (25, 28).
F	Interpret the spirometry when using a race-neutral equation for the calculation of the predicted values	Using GLI global, the same absolute spirometry values result in z-scores for the FEV1 and FVC that are lower than −1.645 and <80% predicted, suggesting restriction. For the patient with dyspnea on exertion, this may prompt further evaluation, which may or may not be helpful depending on whether the patient has a diagnosable and treatable cause of dyspnea.
G	The GLI neutral equations were developed by combining all the data used to develop the race-specific GLI 2012 equations	The GLI 2022 global equations are considered race-neutral because race is not needed to apply the reference equations to patients. They were developed by combining the race-specific GLI 2012 equations using inverse weighting for each group, such that the equations developed from groups with smaller sample sizes (i.e., the African American and North and South East Asian groups) contribute more to the final equation than if the equation were derived without weighting (31). The GLI 2012 “other” equation is another race-neutral equation. It was developed from the same GLI 2012 data but without the inverse weighting. Race-neutral equations perform as well as or better than race-specific equations for predicting symptoms, hospitalizations, and mortality (32–34).
H	Example 2: what is the effect of using GLI African American compared with GLI neutral in terms of predicted values, z-score, and percent predicted for a given FEV1?	For the same FEV1 absolute value, the predicted value will be lower using GLI African American, and therefore the percent predicted and z-score will be higher. Using GLI global, the predicted value will be higher, and therefore the percent predicted and z-score will be lower for the same FEV1 absolute value. If a facilitator or learner wants to review the predicted values, z-score, and percent predicted for a specific patient or specific set of PFT results using different equations, they can use the online GLI calculator: https://gli-calculator.ersnet.org/.
H	What is the impact of using a race-specific or race-neutral equation on decisions about thoracic surgical resection for lung cancer?	Given that the use of GLI global rather than race-specific equations will result in the FEV1 of Black and Asian people being calculated as lower precent predicted and z-scores, this may result in fewer Black and Asian people being considered for more extensive surgical resection of lung cancer because the guidelines on decision-making for thoracic surgical resection use an FEV1 cutoff point. In fact, the relevant value is the predicted postoperative FEV1, but this discussion is typically not needed to illustrate the point. Additionally, surgeons will also consider the patient’s functional state and may request additional testing, such as a quantitative perfusion scan of the lungs and/or a cardiopulmonary exercise test. These details can be provided for more advanced learners if relevant.
I	Example 3: what is the effect of using GLI African American compared with GLI global in terms of predicted values, z-score and percent predicted for a given FVC?	As with the FEV1 in example 2, for the same FVC absolute value, the predicted value will be lower using GLI African American, and therefore the percent predicted and z-score will be higher. Using GLI global, the predicted value will be higher, and therefore the percent predicted and z-score will be lower for the same FVC absolute value.
I	What is the impact of using a race-specific or race neutral equation on decisions about BPAP prescription for patients with ALS?	For a patient with an FVC close to 50% predicted, if they are Black or Asian, they will meet criteria for nocturnal BPAP sooner using GLI global, whereas, for a White patient, GLI global will make their lung function appear better than using GLI White and result in them receiving BPAP later in their disease course.
J	Uses of PFTs	PFT values are used to classify severity of lung disease, patient eligibility for therapies, as well as decisions about disability. The choice of race-neutral or race-specific equation is not straightforward, as the effect of the equation chosen may be positive or negative depending on the individual’s or group’s race and the scenario.
K	Summarize main points	Spirometry interpretation is based on the use of reference range equations to categorize results as normal or abnormal. To use these equations, we need to know the patient’s age, sex, and height. We have also been using race and ethnicity in these equations because of the finding that Black and Asian people have lower lung function than White people (point to series of bell curves). However, race is a social construct, so it is likely that other factors—social determinants of health—are the cause of the observed differences in lung function (point to the list of additional factors associated with lung function). The latest recommendation from the American Thoracic Society is to use a race-neutral equation; however, PFT laboratories can choose which reference equations to use, and this choice can have important clinical and financial implications for patients because it can significantly change the predicted values of the FEV1 and FVC and therefore the percent predicted and z-scores for an individual.

Definition of abbreviations: ALS = amyotrophic lateral sclerosis; BMI = body mass index; BPAP = bilevel positive airway pressure; eGFR = estimated glomerular filtration rate; FEV₁ = forced expiratory volume in 1 second; FVC = forced vital capacity; GLI = Global Lung Function Initiative; PFT = pulmonary function test.

What Is the Content and the Approach?

Before the session, we prepare the board (Figure 1A). We start the session with one of the most common reasons for spirometry: a patient with dyspnea. We frame the discussion by saying the goal of investigations is to diagnose the underlying disease so we can provide treatment and determine the cause of the disease, pointing to where these are written on the right side of the board.

Figure 1.

Board work for teaching racial categorization in reference equations used to interpret spirometry. (A) Illustration of the outline for the session that should be on the board before starting the session. (B–K) Building on this sequentially following the outline of the session provided in What Is the Content and Table 1. ALS = amyotrophic lateral sclerosis; BPAP = bilevel positive airway pressure; FEV₁ = forced expiratory volume in 1 second; FVC = forced vital capacity; GLI = Global Lung Function Initiative; PFT = pulmonary function test.

We invite the learners to interpret the spirometry results first individually and then in groups. While they are talking, we listen to their conversations, circulating as needed to hear their conversation. We tell a group that has correctly interpreted the spirometry as normal that they are correct and ask if they are willing to share their reasoning before inviting them to do so.

Next, we ask the whole group where predicted values for lung function come from. What demographic inputs into the spirometry software are needed to generate predicted values for the individual? We write correct answers on the left of the board (Figure 1B). We add that predicted or reference values come from measuring the lung function of a large group of people thought to have healthy lungs, and that the Global Lung Function Initiative (GLI) 2012 is the most recent, largest dataset for reference equations. When we write “race/ethnicity,” we state that we will discuss this in more detail.

During this discussion, typically learners suggest weight and gender. We acknowledge that very low and very high weights are associated with pathologically reduced lung function (8). However, we explain that, because the addition of body mass index did not add predictive value to reference equations, they are not included (9). We also explain that sex assigned at birth is used rather than gender identity because of ample data showing the association between sex and lung function. Though it is likely that gender-affirming therapies affect lung function, the magnitude probably depends on therapeutic timing and duration and there are few data available about this yet.

Then we explain that lung function, like most physiologic variables, follows a normal distribution, pointing to the bell curve on the board. The predicted value is neither best nor ideal. It is that population’s average value defined by age, sex, and height (one can draw a demonstrative line through the center of the bell curve). The normal distribution is symmetric around this point. We interpret lung function as normal when the values are above a defined cutoff point. We ask the learners to identify this cutoff point. We use this discussion to highlight the shift from percent predicted to z-scores in interpreting spirometry. We explain that, although 80% predicted is a convenient heuristic, referencing lung function to the expected middle value is without a physiologic basis and results in the proportion of the reference population below the threshold varying with sex, age, and height. In contrast, z-scores reflect the spread of values across the reference population, and therefore the proportion of the reference population below the threshold remains constant. We explain that, in medicine and science, we often define 95% of a population as normal and accept that this approach will result in 5% of normal values being considered abnormal. For lung function, we are most concerned about defining the lower limit of normal, so we use a z-score of −1.645 (draw a line on the board on the bell curve here; Figure 1C) as the cutoff point for defining reduced lung function (10). Z-scores indicate how many standard deviations individual values lie away from the mean. Five percent of values lie further than −1.645 standard deviations from the mean (we use our hand to indicate left, away from the line representing −1.645).

We then say that race/ethnicity is on the list of demographic variables because many studies conducted over a period of more than a century found differences in average forced expiratory volume in 1 second (FEV₁) and forced vital capacity (FVC) between racial groups. We are going to focus on the GLI 2012 data set because this combined many existing data sets to generate one of the largest, most diverse spirometry data sets and was the prior recommended standard for interpretation (11). As in prior studies, GLI 2012 data showed that, on average, Asian people’s FEV₁ and FVC are lower than White people’s after adjustment for age, height, and sex, and that Black people’s FEV₁ and FVC are lower than both, resulting in the development of race- and ethnicity-specific equations (12). So, for any given age, sex, and height group, GLI 2012 provided White, North East Asian, South East Asian, and Black equations. We illustrate this by drawing a second bell curve to the left of the first one in a different color (Figure 1D). We label the bell curve on the right “Caucasian” and the one on the left “African American” and state that the curves for the North and South East Asian groups are in between. We make the point that we typically use the terms White and Black, but are using the words “Caucasian” and “African American” here because those were the terms used in the GLI 2012 paper and the African American data set included only Black Americans.

At this point, we sometimes contextualize spirometry inside the larger trend in regard to the use of race in clinical algorithms. We ask learners to volunteer other examples in which mean differences in a clinical parameter are observed across racial or ethnic groups. We distinguish between suggestions for which there has never been a separate interpretative scheme, like blood pressure or white blood cell count, and those for which there has been, like dementia screening or renal function (13, 14). It is our practice to further explore renal function as an example. We explain that the use of race in the estimated glomerular filtration rate algorithm aligned with a stereotype about muscularity in Black individuals and was based on findings from a relatively small number of Black participants, and that the results regarding muscle mass were not replicated in larger trials (15). We highlight how this interpretative scheme led to decreased transplant priority for Black patients and delayed referral to specialty care (16). We then inform learners that similar evolution in thinking has led to the removal of race from algorithms for cardiovascular event risk assessment and vaginal birth after caesarean section, as well as estimated glomerular filtration rate (14).

Returning to the question of spirometry, we ask the learners for their thoughts about this finding. What are some possible explanations for the finding that Black and Asian people have lower lung function than White people? We ask the learners to talk in small groups about this for a few minutes before soliciting suggestions. During this discussion, we emphasize that race is a social rather than a biological construct. We elicit from the learners additional factors that are associated with lung function and write them on the board (Figure 1E) (17–23). We share that, in the United States, many of these factors are experienced differently across racial and/or ethnic groups. For example, Black, Asian, and Hispanic residents experience higher levels of pollution than White residents on average, even after controlling for differences in household income (24, 25). We also share that Black individuals have lower income than non-Hispanic White people at the same level of educational attainment, even among physicians (26, 27). Instructors are encouraged to note the linkages between discriminatory policies and these outcomes. For instance, a simple description of the dynamics around housing and pollution could discuss how legal segregation and practices like redlining concentrated minoritized populations in areas that were perceived as less desirable (25). More complex discussion can invoke evidence for racial bias in historic city planning (28, 29). Social determinants or drivers of health like income, pollution, and neighborhood facilities are likely major contributors to the differences found in lung function between racial groups (30).

We then explain that our example’s spirometry results used the GLI 2012 African American reference equation (add heading above the spirometry results to show this) and then add the results using the GLI 2022 global (i.e., neutral) reference equation (Figure 1F). We ask the learners how they would interpret the spirometry using the neutral equation. What further testing would they recommend? We again encourage small group discussion and identify a group correctly describing results as abnormal and suggestive of restriction. When conversation has quieted, we invite the group we identified to share. We review that, when using the GLI 2022 global equation, this patient’s spirometry would be interpreted as suggesting restriction and typically prompt further evaluation. Therefore, if the patient has an interstitial lung disease, the use of the neutral equation is likely helpful. However, if the patient has reduced lung function due to frequent childhood respiratory infections and exposure to high levels of pollution, a spirometry interpretation suggestive of restriction may not be helpful: the patient will spend time and money on an evaluation that cannot improve his health. From a population perspective, the use of a race-neutral equation may help to reveal the effects of social drivers of health on lung function that were previously masked by “race.”

We explain the GLI 2022 global equations are considered race-neutral because patient race is not needed to apply these equations. They were developed by combining the race-specific GLI 2012 equations (31). We draw a differently colored, wider bell curve atop the others on the board to illustrate (Figure 1G). We share that a number of studies have shown that race-neutral equations perform better than or equal to race-specific equations in terms of the correlation of spirometry values with patient symptoms, hospitalization, and mortality (32–34). We also note that the predicted FEV₁/FVC ratio does not change as much between equations as the FEV₁ and FVC. This means that the identification of airflow obstruction, one of the main uses of spirometry, is minimally impacted.

Next, we present another example: considering surgical eligibility for a patient with early-stage non–small-cell lung cancer. We explain that FEV₁ is the key spirometric value used in thoracic surgery. We ask learners to state if the FEV₁ percent predicted and z-score will be higher using the GLI African American or GLI global equation. As they think, we add the example to the board and then fill the boxes with arrows pointing up or down using learners’ input (Figure 1H). We ask what impact the reference equation selection will have on decisions about surgical resection. Given that use of GLI global will result in the FEV₁ of Black and Asian people being calculated as lower precent predicted and with lower z-scores, will more or fewer be considered for extensive lung cancer surgical resection? Potentially fewer (35).

Our third example involves Center for Medicare and Medicaid Services criteria for prescribing nocturnal bilevel positive airway pressure for a patient with amyotrophic lateral sclerosis, which prolongs survival (36, 37). We explain that the key spirometric criterion here is an FVC <50% predicted. We again add the example to the board with arrows based on learner input (Figure 1I). We ask about the impact on decisions regarding bilevel positive airway pressure prescription. Among those whose FVC is near the 50% threshold, Black and Asian patients will meet the criteria sooner using GLI global than GLI 2012, whereas, for White patients, the opposite would be true.

We emphasize that the choice of reference equations has many effects, including the classification of pulmonary disease severity, eligibility for therapies such as lung transplant, and decisions about disability. We write these on the righthand side of the board (Figure 1J) and make the point that the choice of race-neutral or race-specific equation is not straightforward. The effect of the equation chosen may vary depending on the individual’s race and the scenario.

We summarize the content (Figure 1K) as follows, circling relevant material: spirometry interpretation is based on applying reference equations to categorize results as normal or abnormal. These equations require knowing the patient’s age, sex, and height. Even though we previously used race in these equations because Black and Asian people have lower average lung function than White people, this does not make sense genetically. Because race is a social construct, social drivers of health are the likely cause of the observed differences. The latest recommendation from the American Thoracic Society is to use a race-neutral equation (38). However, pulmonary function testing laboratories can choose which reference equations to use. This choice can have important clinical and financial implications for patients by significantly changing the predicted values of the FEV₁ and FVC, and it is not clear-cut which equations are best to use universally. As in our first example, Black patients with interstitial lung disease might benefit from using the race-neutral equation, but, if they have lung cancer (i.e., second example), the race-specific equation is more likely to lead to surgical management, which is preferred when possible. When learners ask what to do in clinical practice, we encourage them to consider pulmonary function test results as just one data point in their decision-making, work out which equation was used for interpretation, and consider inputting values into the online GLI calculator to understand different possibilities (https://gli-calculator.ersnet.org/).

Why Is This the Approach?

There are four important parts to our approach: 1) address the paradigm shift in medicine away from race-based and toward race-conscious thinking, 2) use evidence-based approaches to engage participants in effortful learning, 3) create a psychologically safe learning environment, and 4) acknowledge uncertainty about the path forward. Why is this the approach?

The paradigm shift of moving from race-based to race-conscious medicine is well described by Cerdeña and coworkers (39). In race-based medicine, race is poorly defined and inferred to have biological significance. This assumption perpetuates race-based stereotypes, such as Black bodies being inherently diseased and inferior to White bodies. In race-conscious medicine, we acknowledge that race theory was developed to justify contemporaneous social inequities (40). Understanding that race is a social construct impacts research, education, and clinical practice because, in this paradigm, racial health inequities can no longer be ascribed to innate differences. Instead, racial health inequities are the mutable consequences of structural racism.

It is well accepted in cognitive psychology that learning requires the learner to expend effort (41–43). Our questions are intended to engage learners in generation: we want them to expend effort attempting to generate the correct answers they may not know beforehand (44). We often use a “think-pair-share” approach, asking learners to consider the question individually, then discuss in small groups (or pairs) before sharing their answers with the group. Think-pair-share is a well-established learning strategy among educators (45, 46) to engage participants in effortful learning and foster a collaborative environment, allowing participants to learn from the group’s diverse perspectives (45, 47). Some educators ask participants to record thoughts before talking in pairs. We walk around during the small group discussions to assess the learners’ understanding of the material. We can then provide input, identify the group that will share in a psychologically safe fashion, and adjust our comments to the collective level of understanding.

The educational benefit of the “share” may be lost if only a few individuals share or if learners are worried about talking in front of the whole group and/or the facilitator (48). Anxiety and other negative emotions are known to decrease learning and test performance (49, 50). The simplest of many explanations for this phenomenon views negative emotions as contributing to extraneous cognitive load and thereby reducing learners’ processing capacity available for the task at hand. We hope to minimize the extraneous cognitive load of the share portion by asking if groups are willing to share and offering the safety of preemptive assurance they are correct. Both steps hopefully reduce the learners’ anxiety. We also try to invite a different small group to share for each question to allow the large group to benefit from the diverse perspectives the small groups present. Psychological safety is particularly important for this topic because of the social and political tensions related to discussing race, as we discuss in the challenges section.

We chose this session’s examples to illustrate the uncertainty about the optimal approach to spirometry interpretation. They illustrate the opportunities and challenges of using a race-neutral equation. Explicitly acknowledging uncertainty serves several purposes: it contributes to psychological safety, highlights the limitations of binary diagnostic approaches, and aims to stimulate creativity in thinking about interpretative approaches.

What Challenges in Teaching the Material Should Be Anticipated?

The most important challenges reflect the potentially contentious nature of racial discourse in American society. This has implications for all parties involved. Underrepresented minority (URM) learners may have their sense of belonging threatened if colleagues think their comments fit the mold of the “angry” minority or if their colleagues’ comments signal something negative about broader attitudes about race. Learners with a majority identity might have anxiety about being perceived as racist. Educators may be concerned about alienating either group—or both. In approximately half of the United States, they may additionally be concerned about local regulations restricting the discussion of topics regarding race and historic inequities (51). Insofar as these concerns inhibit any participant’s best performance, we find that literature originally developed around URM learners is broadly applicable (52).

Educators should attempt to defuse latent racial tension and encourage full participation. One highly successful methodology for transforming tensions into learning opportunities is intergroup dialog. Borrowing from this technique, we encourage active listening (i.e., to understand rather than to debate), normalizing disagreement, and, if sharing personal views, speaking from one’s own experience rather than that of others (53). Sequencing of “personal–institutional” and “concrete–abstract” content has been helpful for us in addressing resistant learners (53). Whereas a participant might hesitate to address larger questions of social inequity or the role of spirometry, they might be more likely to empathize with the particular struggle of Black workers at an asbestos-installation company that, for more than half a decade, faced a higher standard than White workers when making personal disability claims. We observe participants’ voice and body language for signs of frustration or disengagement, including lack of eye contact, excessive personal device use, or turning away from the group.

Second, it is helpful to consider how stereotype threat can manifest as potential barriers for specific groups: URM learners concerned about appearing too radical and learners from majority groups concerned about appearing racist. Stereotype threat is the fear of fulfilling negative stereotypes about one’s group and that maltreatment will plausibly result (52, 54). This fear adds extraneous cognitive load, impairing performance. One example of the general concern about appearing racist arises when a learner finds value in race-specific interpretation. Whether this belief is based on prior literature or concern about a specific case in which race-specific approaches may improve access for Black or Asian individuals, the unifying theme is fear of appearing retrogressive. An educator’s choices may emphasize or alleviate the learner’s anxiety. We recommend reframing and share a number of strategies below, many of which were first developed to reduce stereotype threat (55).

The first act of reframing, which complements all others, is the learners’ understanding of racism itself. Learners often make the maladaptive assumption that shortcomings are primarily a reflection of fundamental personal deficiencies (56). Popular discourse about race can tend to reinforce this tendency. For this reason, we share that the best current understanding does not view racism as an essential part of personality, but rather a descriptor of discrete behaviors (57). We encourage learners that even racist thoughts or actions are single events that can be overcome through persistent practice. Minimizing the extent to which anyone sees the session as a reflection of fundamental character reduces defensiveness. As elsewhere in medicine, we achieve the best results through a growth mindset.

In the general case, reframing means that potentially threatening contexts are considered in new, nonthreatening terms, as when relieving anxiety from a test by sharing that it does not reflect fundamental differences in ability. Our goal in reframing is to create the imaginative space for similarly nonthreatening interpretations in the context of racial dialogue. In the case of spirometry, one of the arguments for race-specific equations is to prioritize precision and avoid worsening existing health disparities. One can explain that it is mathematically impossible to fulfill all notions of fairness simultaneously (58). Achieving different fairness criteria may therefore imply including or excluding different variables. To exemplify this process and stimulate creativity in imagining alternative approaches to spirometry interpretation, we often discuss the approach to diagnosing hypertension as prioritizing clinical utility. This uses a single threshold that predicts adverse cardiovascular events despite the finding that, on average, Black Americans have higher blood pressure than White Americans (59). Equally, a public health perspective seeking to understand the respiratory effects of pollution or poor nutrition might include these variables in an analysis to estimate their impact size. Race could be used in a similar manner, illustrating that the use of race is not inherently racist.

This exercise segues from reframing into a second important strategy to prevent stereotype threat: blurring group boundaries (55). Regardless of the specific examples or thought exercises used, the goal is to reframe in a way that activates shared identities and blurs group boundaries. Our suggestion to explore the impossibility of satisfying all fairness criteria is designed to highlight actual roles our learners may have currently or imminently assume. Inviting their participation to name how including or excluding a particular variable may be helpful in some contexts but putatively harmful in others can strengthen the effects of this exercise. Engaging their various roles as scientists, epidemiologists, and clinicians expands the number of identities available to them during the session and may reduce their dependence on a potentially threatened identity.

Another type of reframing we employ is to encourage learners to think about the process of scientific discovery rather than judging the outcome alone. This means considering how scientific understanding often does not develop smoothly, but rather with uncertainty and missteps along the way (60). Potential examples include one’s own intellectual journey on the topic of race in spirometry and Juan Carlos Finlay’s work on yellow fever (61). He was among the first to recognize mosquitos as the critical disease vector. However, unsuccessful trials that attempted to directly demonstrate spread through mosquitoes led to an incorrect denunciation of him as irrational. When he later shifted to epidemiologic studies associating mosquito control with yellow fever incidence, he was instead praised. The evolving perception from fringe theorist to visionary public health advocate for the same intellectual position demonstrates the irregular trajectory of science. We stress that the work of understanding the determinants of human lung function is similarly incomplete. We normalize that error is inevitable in science and destigmatize exploring the utilities of race-specific interpretation, even as we make clear that the latest guideline and much current evidence does not support it.

We are sometimes asked directly about “racism” in the history of spirometry. Facilitators should explain that the work of the best-known commentators—Thomas Jefferson, Frederick Hoffman, and Samuel Cartwright—represents a broader social sentiment of their time (62–64). We generally do not discuss the specifics of each author’s claims, as this recirculates harmful ideologies and may impair some trainees’ learning. Instead, we summarize the point that these authors employed lung function data to justify the status quo of subjugating Black peoples. We reiterate that this was originally racial categorization’s central function, and that these arguments were deployed as central justifications for slavery (65). We distinguish this intentional, explicit bias from the concepts of implicit bias and structural racism, sharing that the work of later authors, who continued to use race-specific distinctions, was shaped by the implicit biases of their social contexts. When race-specific algorithms contribute to inequitable clinical care, the algorithms are examples of structural racism. We use this contrast to highlight an important quality of systemic racism: it produces targeted harm against one group even without ill intent by any individual. We prioritize acknowledging the harms caused by racist beliefs over denouncing racism in the abstract sense.

A final important challenge is the facilitator’s own confidence with the material. Individuals with more experience addressing health equity have more resources to broach this material, just as institutions with broader engagement on equity will have greater communal reservoirs of good will to draw from. Nonetheless, a lack of specific expertise should not be viewed as prohibitive. Instead, we advocate for applying the principles of cultural humility (66). This means facilitators should admit their limitations in discussing race and its role in health outcomes: invite learners to correct your misstatements and frame the session as a discussion during which all members of the group can learn from each other. Stress this as an area of lifelong learning for all involved. Emphasize that it is unhelpful to feel bad for what one has not had the opportunity to learn. Modeling this vulnerability may be more broadly corrective of the toxic pressure toward perfectionism in medicine (67).

Teaching the most current understanding of spirometry is a microcosm of the challenge of creating a more inclusive, culturally robust healthcare system. In its particulars, we can perceive many anxieties of American racial discourse. However, solutions broadly mirror several best practices for fostering inclusivity: lessons from intergroup dialogues, cultural humility, and reframing as a strategy to address stereotype threat are all helpful. In this way, much like the new recommendations themselves, solutions that were originally tailored to assist narrow segments of the population can prove to offer universal benefits (68).