Consistency and inconsistency with which sociodemographic variables are associated with performance on medical school...

The tools and processes used to select applicants to medical school and residency training play a critical role in determining the future of healthcare. As selection strategies evolve to keep up with the competencies expected of physicians and the social accountability mandates of medical education programs, it is increasingly imperative that we develop awareness of how equity is influenced by how admissions decisions are made. This study was, thus, conducted to explore the consistency with which sociodemographic variables are associated with scores on academic and non-academic medical school selection tools.

Retrospective cohort study of 6 successive application cycles (2016-2021) undertaken at the University of British Columbia's Undergraduate MD Program. Six sociodemographic variables were gathered or constructed from data available in the program's admissions database: Applicants' age, gender, high school location, self-identified disability, Indigeneity, and an educational-occupational index that reflects socioeconomic status. Subgroup differences were assessed for each cohort on each sociodemographic variable for 5 admissions tools: Grade point average, Medical College Admissions Test (MCAT), Non-academic activity assessment, Multiple Mini-Interviews, and a Remote and Rural Suitability Score.

N = 14,781 applicants were included. Although patterns emerged, variability across cohort was also prominent. Large and consistent differences were observed between age groups and between Indigenous and non-Indigenous applicants for both academic measures (GPA: d = 0.96 and 0.63 for age and Indigeneity, respectively; MCAT: d = 0.84 and 0.86, respectively). Other associations were less robust.

These data demonstrate the importance of taking the entire admissions system into account when making policy decisions rather than simply debating the value of tools independent of one another. Further, the data reveal the need to treat quality assurance efforts in a longitudinal manner rather than risk being misled by assuming any given cohort year to be representative of more general patterns.