Multivariate analysis of follow-up physiological data recorded by cardiac implantable devices

New cardiac implantable devices (IDs) allow the acquisition of an increasing amount of data relative to the patient's activity. A quantitative analysis of these data can be of particular interest to optimize the stimulation therapy and to improve patient follow-up. This work presents a method to: i) evaluate the information content of the ID memory data; ii) define synthetic indexes that would summarize these data and ease their interpretation and iii) characterize and compare different patient populations. The proposed approach is based on a coding stage allowing the representation of time-varying data, followed by a multivariate analysis stage, based on principal component analysis (PCA). Results show that the initial 38 variables obtained from the IDs memory contain pertinent information that can be synthesized with a low number of indexes (the first four factorial axes of the PCA represent 70% of the total variance) and that these indexes can be useful to follow the evolution of a given patient's state.