The one-to-one apprenticeship in many residency programs, while providing direct supervision from an assigned expert, lacks timely and diverse feedback, especially from other physicians with unique clinical tendencies. This often contributes to subpar training for high-stakes tasks with unstructured solutions, further diminishing education quality and patient safety. This work specifically explores the case of contouring — delineating tumor and healthy tissues — that is often regarded as the weakest link in radiotherapy treatment planning due to pervasive errors that lead to detrimental consequences for patient safety. This paper first offers three design goals aimed to prevent gaming, balance between expert consensus and tendencies, and minimize cognitive load. This work then designs and develops iConTutor, a learning platform that not only provides timely feedback, but also presents crowdsourced expert contours as part of distinct feedback mechanisms. A pre- and post-test study with nine residents showed a 32% increase in contouring accuracy, and the participants benefited from iConTutor’s feedback mechanisms in delivering light-touch, skill-based, and awareness-cueing training. The three design goals of this work (implemented as part of feedback strategies in iConTutor) can inform computer-supported learning tools in other healthcare domains that aim to improve apprenticeship training via timely and diverse feedback.
Yarmand et al. (Mon,) studied this question.