Abstract
Classification and regression trees (CART), was applied to develop decision rules to investigate about the conditions for determining a discharge destination in aged stroke patients. Descriptive and functional status data for the 374 patients were collected at the time of discharge. 82.6% of the patients were discharged to home. In the analysis of single variable, the significant differences concerning the discharge destination were observed in the numbers of family, the existence of a spouse living with a patient, and each functional independence measure (FIM) score (18 items) at the time of discharge. The decision tree, which consists of a FIM toilet transfer and the family number, was obtained as a result of the analysis by CART. When the patient was needed a care for toilet transfer, and when the numbers of family was two or less, it was difficult for the patient to discharge to home (rate of discharge to home was 21.7%). CART is considered to be an effective tool in research for classification, prediction, etc, because it is able to handle both a continuous variable and a category variable, and to be understood a result intuitively.
