If VR-based medical training and assessment is to improve patient care and safety (i.e. a genuine health gain), it has to be based on clinically relevant measurement of performance. Metrics on errors are particularly useful for capturing and correcting undesired behaviors before they occur in the operating room. However, translating clinically relevant metrics and errors into meaningful system design is a challenging process. This paper discusses how an existing task and error analysis was translated into the system design of a VR-based training and assessment environment for Ultrasound Guided Regional Anesthesia (UGRA).