Presentation (CTI)

OBJECTIVES: Persistent racial disparities in U.S. healthcare remain a central policy concern. Black mortality has been 17-34% higher than non-Hispanic White mortality since 2018, and minority populations receive worse care across 30-43% of quality measures. A growing literature documents that physician-patient racial concordance improves health outcomes for minority patients, yet minorities remain severely underrepresented in medicine. This paper asks: do physicians who trained alongside more minority peers deliver better outcomes for minority patients later in practice?
METHODS: We link three administrative sources: AAMC cohort data (2014-2023) on medical school racial composition, the Florida Physicians' Workforce Survey containing residency training records for over 115,000 physicians, and Florida AHCA Hospital Inpatient Data (2016-2023) linking patient outcomes to treating physicians. We construct physician-level peer exposure indices capturing horizontal exposure to minority co-residents, cohort composition at entry, and vertical exposure to minority senior residents. A fixed-effects specification regresses patient-level outcomes on peer exposure, a minority patient indicator, and their interaction, controlling for medical school, residency program, hospital, physician, and year fixed effects. To address non-random patient sorting, we instrument for peer exposure using the share of on-duty physicians at a given hospital, weekday, and hour who trained in high-diversity environments.
RESULTS: Across fixed-effects and IV specifications, non-Black physicians who trained alongside more Black peers during residency deliver significantly better outcomes for Black patients. The residency peer exposure index is associated with reduced Black patient mortality, and IV estimates confirm a significant interaction effect. We find parallel effects for senior peer exposure, with non-Black physicians mentored by Black senior residents also showing improved outcomes for Black patients.
CONCLUSIONS: Exploring mechanisms, we find no evidence that peer exposure improves clinical knowledge or resource utilization. Effects appear to operate through trust and communication, as measured by reductions in patients leaving against medical advice.