Prediction Models of Sudden Death from Chronic Chagas Cardiomyopathy

Chagas disease is an endemic disease in Latin America that cause cardiomyopathy in its chronic stage. The symptoms include heart rhythm disturbance that can evolve to sudden cardiac arrest. This paper presents methods for stratifying and predicting the risk of sudden death in chronic Chagas’ cardiomyopathy, retrospectively analyzing the 24 h Holter records of patients. The database includes 60 cases of alive patients and 22 cases that evolved to sudden death. For this study, 13 predictor variables were found from an ECG signal. They were extracted from the heart rate turbulence and heart rate variability parameters in the time domain. These variables were used as input to seven classifier models, these being Support Vector Machine, Multilayer Perceptron, Python’s Keras Library, K-nearest Neighbors, Logistic Regression, Randon Forest, and Decision Trees. The best model (Randon Forest) provided 95% accuracy, 100% sensitivity, and an area under the ROC curve of 95.8%, fitting the assessment of an excellent classifier.