Background <p>Communication skills are central to clinical practice and are typically assessed through Objective Structured Clinical Examinations (OSCEs). Prior research indicates that self-reported questionnaires, simulated-patient evaluations, and expert assessments each offer valuable yet non-interchangeable perspectives on medical students’ communication skills. However, it remains unclear whether these assessment modalities, when collected during the academic year, can predict subsequent OSCE performances and thereby help identify students struggling to meet communication competency standards. This study examined whether self-reported empathy, the ability to recognize others’ emotions, and simulated-patient evaluations of communication skills predict later OSCE performances in medical students.</p> Methods <p>Data were drawn from the ETMED‑L longitudinal open-cohort project at the University of Lausanne. The third- and fifth-year medical students who completed the yearly ETMED‑L online questionnaire and consented to the use of their communication skills evaluations were eligible for the present study. Self-reported empathy was assessed using validated multidimensional instruments and the ability to recognize others’ emotions was measured with a performance-based task. Simulated patients used a standardized form to evaluate communication skills during mandatory formative interviews. OSCE performances were rated by experts using structured communication grids and total exam points. Structural equation modelling was used to examine how self-reported empathy, the ability to recognize others’ emotions, and simulated-patient evaluations independently and jointly predict OSCE performances.</p> Results <p>A sample of 468 third‑year and 399 fifth‑year students were analyzed. In both cohorts, self‑reported empathy, the ability to recognize others’ emotions, and simulated-patient evaluations independently predicted OSCE performances. In combined models, only simulated-patient evaluations remained significant predictors of OSCE performances, with a moderate effect size for the third-year cohort and a small effect size for the fifth-year cohort.</p> Conclusions <p>Simulated-patient evaluations of student communication skills during formative training are more strongly associated with later expert-rated OSCE performances than both self-reported empathy and a performance-based measure of the ability to recognize others’ emotions. These findings underscore the practical value of structured simulated‑patient programs within competency‑based medical curricula, as they may help identify students who have difficulty attaining required communication competencies.</p>