Early specialty preference status among first-year medical students: a three-cohort observational study of role models and academic interest

Background Early in medical school, some students already report a specialty preference while others do not, but the factors associated with early preference reporting are unclear. Using Social Cognitive Career Theory (SCCT) as an interpretive lens, we examined two entry-measurable factors (role model presence and academic-interest importance), adjusting for regional quota admission, and, secondarily, whether depressive and anxiety symptoms were associated with preference status. Methods In a three-cohort observational study at a Japanese national university (2023–2025 first-year classes; participation 63%, 84%, and 89%), students were surveyed early in the first year (T1) and in the second semester (T2). The outcome was T1 specialty preference status (named versus undecided). Binary logistic regression estimated adjusted odds ratios (OR) and average marginal effects, with Firth, nonresponse, outcome-definition, and ten-motive sensitivity analyses; role-model sources were compared post hoc, and T2 responses described preference stability. Results Among 282 students, role model presence (adjusted probability + 17 percentage points, 95% CI 7–28; OR 2.99) and academic-interest importance (+ 7 points per SD, 2–12; OR 1.49) were independently associated with reporting a named field at T1; the latter held after accounting for all ten motives. Regional quota admission was also strongly associated (+ 19 points, 7–31; OR 3.62). Named-field proportions were 100% (21/21) with non-family role models (mostly physicians), 74% with family role models, and 68% with none. Depressive and anxiety symptoms showed no clear association (OR 0.89 and 0.93 per SD in separate models), with intervals compatible with negligible or modest associations in either direction. Only 48% of fields named at T1 were named again at T2. Conclusions Students who reported a career role model or assigned greater importance to academic interest were more likely to report a named field early in the first year; in post hoc analysis, the role-model association was concentrated among non-family role models, and the regional quota association likely reflects self-selection into a policy-defined pathway. These mutually adjusted cross-sectional associations do not establish how preferences form, and many early preferences were tentative. Estimates for depressive and anxiety symptoms were imprecise; an undecided response should not be used as a stand-alone indicator of psychological distress.