The aim of this work is twofold: to illustrate power of unsupervised data analysis approach on routinely collected diagnostic data for coronary heart disease patients and to validate findings against cardiologist’s own patient classification and expert analysis. In this respect emphasis in this work is not on prediction and accuracy but rather on discovering paths to extraction of new insights and/or knowledge of the domain. The work demonstrates the use of unsupervised classification for the partitioning of the database with the aim of amplifying predictability of models describing expert classification, as well as boosting cause-and-effect relationships hidden in data.