Project summary. Health disparities are a matter of grave public health significance1, and producing evidence to make health care more equitable is part of the Agency for Health Care Research and Quality (AHRQ) mission. Racial health disparities have complex etiologies and correlates, but remain when controlling for other social determinants of health2 and patient factors (e.g. treatment refusal3). Residual disparities reflect differences in provider treatment of White and minority patientse.g.4–6. One factor contributing to disparate treatment of minority patients is provider implicit bias—non-conscious biases that alter behavior7. Provider implicit bias predicts subtle behavioral differences in interactions with minority patients, including more anxiety- related words8, more negative affect9, and different nonverbals10. These provider behaviors predict lower patient satisfaction and adherence11, with large health consequences11,12. Despite haste to target implicit bias in disparity reduction interventions, most studies show no impact of implicit bias interventions19. Further, implicit bias is difficult to measure and demonstrates moderate test-retest reliability13. Still, there are public health implications of even weak effects of implicit bias14 when considering the number of people affected. Effects of implicit bias on disparities may be clarified by articulating and examining more complex models of the relationship between implicit bias and provider behavior. This proposal examines intergroup anxiety (anxiety that manifests in interracial interactions in response to negative expectations15) as a mediator of the relationship between implicit bias and provider behavior. It is well known that anxiety affects behavior in the general populatione.g.16, and provider anxiety impairs patient outcomes, such as satisfaction and adherence17,18, but no research has examined the effects of intergroup anxiety on provider behavior. We will test this model in a sample (N=70) of medical students. Participants will each interact with two patient actors—one Black, and one White—to control for race-irrelevant anxiety. To ensure a comprehensive analysis of the innovative association between intergroup anxiety and provider behavior, we propose to measure both constructs at multiple levels. We will assess anxiety through self-report affect and physiology19. We will examine three classes of behavior: verbal (anxiety-related word use), global (warmth), and nonverbal (smiling and eye contact). Medical school is a key window-of-opportunity when biases may be more malleable20, students are accessible, and training is expected. Many medical schools use implicit bias reduction trainings to decrease disparities, but intergroup anxiety may represent a more consistently alterable and easy-to-measure construct. Disparity-reduction trainings based on evidence-based models such as the proposed may have large impacts on health disparities, addressing the AHRQ priority focus of...