Institutional characteristics associated with meeting the proficiency benchmark in Brazil’s National Medical Education Assessment (ENAMED): a nationwide Bayesian hierarchical analysis

Abstract

Abstract Background Brazil’s medical education system has expanded at unprecedented pacing, largely propelled by policies meant to increase the physician workforce and reduce regional inequities. This expansion succeeded in scale, but it reignited an old question: are we expanding training capacity or expanding diplomas? In 2025, Brazil introduced the National Medical Education Assessment Examination (ENAMED), a regulatory instrument grounded in a simple principle: programs should be judged, and effectively ranked, by the proportion of graduating students who meet a defined minimum standard of professional proficiency (knowledge-based), rather than by mean scores alone. Here, we evaluated institutional and structural determinants of ENAMED performance, focusing on the system’s ability to guarantee minimum proficiency across medical schools amid recent expansion patterns. Methods We conducted a nationwide retrospective observational analysis linking ENAMED-2025 results to institutional and regulatory characteristics of medical programs. Our outcome was systemic assurance of minimum proficiency, defined as meeting a regulatory benchmark in which ≥ 60% of graduates are classified as proficient. We fit hierarchical Bayesian models estimating each program’s posterior probability of meeting this benchmark, accounting for course size and geographic clustering (state and municipality). Results ENAMED revealed pronounced heterogeneity in the probability of meeting the minimum proficiency benchmark across institutional segments. Federal and state public medical schools had the highest assurance, followed by community-based institutions. Private programs, especially for-profit schools, showed substantially lower probabilities of meeting the benchmark. Programs created during the most recent waves of expansion and those with larger volumes of authorized enrollments consistently exhibited weaker assurance. Independently, greater enrollment capacity was associated with lower probability of achieving the ≥ 60% proficiency threshold, with the steepest gradient observed among for-profit institutions. Conclusions ENAMED’s signal is strongly shaped by institutional design and conditions under which expansion occurred, far more than by chance variation. By privileging the share of graduates who reach a minimum proficiency standard, ENAMED shifts regulation from “average performance” to “minimum guarantees,” offering a more defensible lens for accountability in medical education. Its promise, however, depends on policy responses that strengthen academic governance, clinical training capacity, and faculty support, rather than short-term, reactive strategies aimed at gaming exam performance.