Assessing the semantic content of clinical case presentations: studies of reliability and concurrent validity