Predictive Models for Risk Assessment of Worsening Events in Chronic Heart Failure Patients

Maria Carmela Groccia1, Danilo Lofaro1, Domenico Conforti1, Angela Sciacqua2
1Università della Calabria, 2Università “Magna Graecia”


Abstract

This work aims at developing and assessing a machine learning based Knowledge Discovery (KD) task for risk prediction of major cardiovascular worsening events in chronic heart failure patients. We analysed clinical data from 50 patients with chronic heart failure collected at the Cardiovascular Diseases Division, Department of Experimental and Clinical Medicine, University "Magna Graecia" of Catanzaro, Italy. For each patient, personal data, different vital and clinical parameters and the presence of cardiovascular worsening events have been stored every three months per two years. We defined the KD analysis as a predictive task stated as supervised binary classification problem; every instance is represented by the general patients’ information and the value of the clinical parameters measured during a visit. The class label was defined based on the occurrence or not of cardiovascular worsening events between two consecutive visits. A temporal weighting strategy was applied to take into account the temporality of the worsening events. Several machine learning algorithms were applied to collected data obtaining different predictive models: Linear/Radial Support Vector Machine (SVM), Decision Trees, Naïve Bayes and Neural Networks. Parameters tuning has been optimized basing of performance on a 5-fold cross-validation. Models performance have been evaluated in term of area under the ROC curve (AUC), and Linear SVM got the best performing predictive model. In conclusion, the implemented KD task have shown to be a reliable tool for support cardiologists for risk predictions of major cardiovascular worsening events.