In a smart hospital, effective decision supports are useful for medical diagnosis. Recent advances in the field of data mining, pervasive computing and other computing methods are ready to meet this kind of challenges. However, few techniques can be gracefully adopted for generating accurate and reliable as well as biologically interpretable rules. The objective of this paper is to introduce a novel method for classifying coronary artery disease dataset based on the principle of decision trees. We extend classical decision tree building algorithms to handle data sets with Multivariate in nature. Extensive experiments have been conducted which shows that the resulting classifiers are more accurate than the existing classifiers. The performance of the algorithm is evaluated with the coronary artery disease (CAD) data sets taken from University California Irvine (UCI). © 2012 Springer-Verlag.