Entrustable professional activities (EPAs) are used to support clinical skills training by providing a scaffold for practice and feedback. The aim of this study was to evaluate how written feedback provided for EPAs supports medical students' learning. The written feedback for two EPAs (admit or consult patient and discharge summary) was analysed using a mixed methods approach. Content analysis determined the types of feedback received. Qualitative analysis identified themes related to the feedback quality. Hattie and Timperley's feedback model that provides a framework for effective feedback was used as an interpretive lens for our analysis and integrated findings. Written feedback was provided for 89.5% of 997 EPAs. The proportions of feedback types to support learning were as follows: feedback related to what was done well (75.6%), areas for improvement (27.7%) and next steps (17.4%). Only 10% of EPAs received all three types of feedback. EPAs completed independently were more likely to receive any feedback. Feedback for supervised EPAs was more likely to include areas for improvement or next steps. Qualitative analysis identified two themes: effective feedback for present and future performance with detailed description of performance or targeted advice, and lack of specific guidance for improvement with affirmative statements or non-specific directions. EPAs most often elicited feedback about what is done well without areas or strategies for improvement. Descriptive feedback and tailored advice are ideal, but uncommon. Students and supervisors need guidance on how to optimise feedback through instruction and EPA form design.