Implicit racial/ethnic bias and diagnostic decisions

Despite considerable research activity on racial/ethnic disparities in health care, gaps exist in the understanding of factors on the micro-level, i.e., those surfacing during provider-patient interaction. This study considers two such potential factors: (1) providers' implicit, i.e., non-conscious, racial/ethnic biases and (2) stress. Scholarship in cognitive social psychology indicates that implicit racial/ethnic biases, which are learned through early socialization, play an important role in everyday interaction, and that their effects are accentuated under stress. Applied to health care contexts, this theoretical framework yields two hypotheses: (1) Implicit racial/ethnic biases, when activated, influence medical decisions in ways that disadvantage minority patients and (2) The effects of implicit racial/ethnic biases on medical decisions are greater when providers are under high stress than when they are under low stress. This paper presents a pilot experimental study evaluating the proposed hypotheses. The study was conducted online with 44 primary care physicians. Physicians were randomly assigned to 4 racial/ethnic conditions (Black, Hispanic, White, and neutral) and two stress conditions (high or low stress). To activate implicit biases, participants were shown very brief flashes of words representing racial/ethnic categories, i.e., black, white, Hispanic, or race-neutral words. Very brief presentation ensured that activated biases were implicit, not explicit. Afterwards, physicians read a description of a hypothetical patient with chronic unstable angina and answered questions about how they would diagnose the patient. Finally, physicians reported their practice and demographic characteristics and were debriefed. Preliminary hypotheses testing was performed by estimating multivariate general linear regression models. Dependent variables were physicians' evaluations of the likelihood that the patient suffered from angina and coronary heart disease (CHD). The experimental factors (assignment to the racial/ethnic group and assignment to a high or low stress group) served as the main independent variables. Interaction effects between the main experimental factors were tested. The models controlled for socio-demographic and practice characteristics. The results yielded initial support for the hypotheses. Specifically, physicians in the Black condition gave lower evaluations of the likelihood that the patient had CHD compared to physicians in the neutral condition. They also evaluated the likelihood that the patient's chest pain was angina lower compared to physicians in White condition. In addition, the difference between evaluations of angina in white and Hispanic condition was larger when physicians were under high stress compared to when they were under low stress. Limitations and further directions are discussed.