This research used a text data mining technique to extract useful information from text data within Electronic Medical Records. Although text data provide a complete account of a patient's information, they are not being fully utilized. Such relevant information as laboratory results and remarks made by doctors and nurses is not always considered. Knowledge concerning the condition and treatment of patients has been determined in a twofold manner: a text data mining technique identified the relations between feature vocabularies seen in past hospital patient records accumulated on the University of Miyazaki Hospital's Electronic Medical Record, and extractions were made from those relations. The analysis discovered vocabularies relating to proper treatment methods and concisely summarized their extracts from hospital patient records. Important vocabularies that characterize each hospital patient record were also revealed. The results of this research will contribute to nursing work evaluation and education.