Clustering Cardiac Rehabilitation Data: A Preliminary Study

Abstract

In this paper, a clustering framework is used on cardiac rehabilitation data to discover meaningful patterns. The data were collected in three phases. K-means clusters were generated and evaluated for stability. Visual assessment of the clusters using PCA plots was also done. A scoring system was developed to quantify improvement in the patients’ health across the three phases. With the scores, association and correlation measures were employed to assess the meaningfulness of the clusters. Two distinct clusters were found and they were shown to have moderate clinical association (Cramer’s V score=0.27) with the improvement scores.