Over the last decade, research on racial differences in the quality of the patient–physician relationship has yielded mounting evidence of lower levels of trust,1 satisfaction2 and perceived partnership2 in race-discordant patient–physician relationships. Evidence linking race discordance to technical aspects of care and health outcomes is limited and results are mixed.3 However, continued study of the role of race discordance in perpetuating or mitigating health and health care disparities is important because most relationships between ethnic minority patients and their clinicians are race-discordant.4

More recent work focuses on clarifying mechanisms of poorer patient experiences in racially discordant patient–physician interactions. Key mechanisms include poorer quality communication, especially in affective domains,2 possibly due to implicit bias or stereotyping of patients that occurs during medical interactions.5 Although several intervention strategies have been proposed to reduce the impact of bias on health care disparities,6 to date, relatively few studies have used rigorous research methods to examine the effectiveness of these strategies. Interventions with potential for alleviating the effects of bias include programs designed to enhance participatory decision-making between African-American and poor patients and their physicians7 and an intervention that used a values-affirmation exercise before medical visits to reduce the impact of stereotype threat on African-American patients.8

The article by Penner and colleagues4 in this issue of JGIM investigates the effect of an intervention based on the common ingroup identity model on physician and patient perceptions of a racially discordant medical encounter, patient trust and patient adherence. Social psychologists have postulated that categorization into social groups is a process fundamental to intergroup bias and have targeted it as a starting point for improving intergroup relations. The goal of the common ingroup identity model is to reduce bias by changing sensitivities regarding intergroup boundaries and redefining who is seen as an ingroup member.9 By encouraging members of different groups to conceive of themselves as a single, more inclusive superordinate group via recategorization, attitudes toward former outgroup members are expected to become more positive through processes involving pro ingroup bias, which would thereby reduce intergroup bias. Use of the common ingroup identity model has consistently demonstrated reduction of intergroup bias and more positive intergroup attitudes across various settings.9 Here, Penner and colleagues apply the common ingroup identity model to address the problem of racial bias and its effect on health care disparities. Although the authors did not observe any differences between the treatment and control groups immediately post-interaction, they found that the intervention had longer term effects on patient trust of physicians and adherence to treatment recommendations. These results hold promise for the future of improving the quality of racially discordant patient–physician interactions and subsequent health outcomes, and like Penner and colleagues, we are cautiously optimistic. Interventions based on the common ingroup identity model can and should be tested with different patient and physician populations, so that we may optimize the potential that this theoretical model has to inform interventions that diminish racial bias in the healthcare setting. In addition, further research in this area has the potential to address intergroup bias in other socially discordant patient–physician relationships.

The prospect of replicating the common ingroup identity intervention in other populations is an exciting one. It prompts us to consider situations in which the results may or may not be consistent with hypothesized effects. This may be helpful as research in this area forges ahead to discover for whom, and under which conditions, this intervention would be most effective. Since all interactions in this study were racially discordant, it is unclear whether the intervention would enhance trust and adherence to a similar extent in racially concordant interactions. Like other interventions based on theoretical models that promote commonality, the intervention may enhance relationships, regardless of race discordance. If so, disparities in racially discordant and concordant visits might persist, or widen. Additionally, patients in this study were African Americans with low income. Is it more or less challenging to reduce intergroup bias when patients and physicians are discordant in more than one dominant social category? Racial identity was not measured in this analysis. We wonder whether the intervention would be more or less successful in a sample of black patients who identify more or less strongly with their race. Might this intervention also mitigate the effects of bias for other groups affected by health disparities, such as obese patients or those with disabilities?

Characteristics of physicians in the study also influence our interpretation of the results and raise questions for future research. Penner and colleagues did not observe any post-intervention physician treatment effects. Participating physicians were mostly non-black minorities from the Asian subcontinent (the rest were white), and all had previous training in patient centeredness and a slight pro-black implicit bias. Despite this previously identified limitation, the intervention eventually worked for patients; therefore, we wonder if it would have had more immediate and marked effects if participating physicians had more pro-white bias. Moreover, how would differences in the overall baseline prevalence of training in cultural competence and patient centeredness, or experience working with diverse patient populations change our ability to detect an effect on the physician level? The randomized design should have minimized differences between intervention and control physicians in most characteristics; however, since many social variables were unmeasured, levels of social concordance on other dimensions or shared identity with patients may have impacted the results.

Application of the common ingroup identity intervention in the healthcare context also raises the notion of power imbalances inherent in patient–physician relationships.10 Do these imbalances contribute to greater intergroup bias in race-discordant relationships? In line with the complex role of collective identities and potential limitations of the common ingroup identity model described by Dovidio and colleagues,9 are white physicians more motivated to protect their collective high status identities as whites and health professionals? At the same time, are African-American patients more motivated to enhance the collective identity of their own racial group? If so, white physicians might prefer a common one-group identity while African-American patients might prefer a dual identity in which their differences from physicians are acknowledged in the context of a superordinate identity such as a team. An intervention that encourages commonality and ignores this motivation for a dual identity might fail to address healthcare disparities. Does this same potential threat exist for other relationship-centered care interventions aiming to equalize power between patients and physicians?

To maximize our knowledge about how theoretically-based interventions as the common ingroup identity model can reduce intergroup bias in healthcare, future research should examine the relationship between these interventions and objective measures of physician and patient behaviors (e.g., verbal and nonverbal communication). In addition to studying intervention effects on patient perceptions, changes in health professionals’ perceptions of individual patients as well as changes in actual clinical decisions should also be examined. Furthermore, long-term follow-up of study participants is important to assess the sustainability of intervention effects. Further examination of these interventions in varied populations should also take into account the most appropriate representation of the common ingroup identity model to improve intergroup relations.9

In conclusion, the application of the common ingroup identity model is a promising strategy to reduce bias and poorer health care experiences in racially discordant patient–physician interactions. This remarkable study is also a testament to the value of transdisciplinary collaboration in developing innovative strategies to answer the most intriguing questions and to solve the tremendous problem of bias in health care.