Pandemic to endemic: a longitudinal multinational study on COVID-19’s effects on graduate medical education

BACKGROUND: This study applied a resilience engineering framework to understand how international graduate medical educations (GME) programs responded to the disruptions caused by the COVID-19 pandemic. Beyond describing early pandemic disruptions, we offer a multi-year, multi-country analysis of how training functions adapted, which elements durably changed, and what program design features may be associated with recovery. METHODS: We conducted a retrospective longitudinal study of all 180 ACGME International (ACGME-I)-accredited programs across seven countries. Annual surveys were completed over three years (Year (Y)1 = Academic Year (AY)2021-22, Y2 = AY2022-23, Y3 = AY2023-24), assessing clinical experiences, educational activities, telemedicine use, and training modifications. Data were analyzed using mixed-effects and generalized linear mixed models, with exact logistic regression for small strata. Subgroup analyses compared specialty, training type, and World Bank income classification. Studying ACGME-I programs provides a common competency-based context, allowing us to describe trajectories of disruption, recovery, and persistence. RESULTS: Of 180 programs, most were residencies (68.3%) and medical specialties (53.3%), with 87.2% located in high-income countries. Clinical experiences were most disrupted in Y1, especially ambulatory continuity clinics (estimated marginal means [EMM] 1.489) and ambulatory rotations (1.571), both of which recovered by Y3 (1.863 and 1.874; p<.001). Inpatient admissions showed similar recovery (1.915→2.120; p<.001), while urgent procedures declined slightly (2.013→1.897; p=.007). Emergency and ICU exposures initially increased (2.242 and 2.342) but returned toward baseline by Y3 (2.232; 2.244; both p > .05). Educational activities were most disrupted for in-person didactics, which nearly ceased in Y1 (0.784) but improved by Y3 (1.386; p<.001), offset by a surge in live virtual conferences (2.821→2.476; p<.001). Elective rotations also declined (1.440→1.782; p<.001), whereas scholarly activity (1.836→1.956; p=.002) and direct observation (1.879→1.954; p=.005) showed slight disruptions and modest increases by Y3, while one-to-one advising (1.970→2.012; p > .05) remained stable. Telemedicine use peaked in Y1 (clinical care 1.020; supervision 0.544) and declined by Y3 (0.720; 0.386; both p<.001) but remained in use across programs. Training extensions declined substantially from 43.2% in Y1 to 23.0% in Y3 (OR = 0.11; p=.002). Differences by specialty and country income suggest heterogeneous recovery pathways rather than a uniform "return to normal." CONCLUSIONS: International GME programs demonstrated resilience, restoring clinical and educational activities, integrating telemedicine, and markedly reducing training extensions. Our longitudinal, multi-country data move the field beyond single-site, early pandemic reports by identifying which disruptions resolved, which practices persisted, and which gaps remained. Because all programs shared competency-based structures, findings suggest that specific design features may support adaptation. These results refine the prevailing "back to business as usual" narrative in international graduate medical education by showing a selective reversion with durable innovations.