Multiscale entropy analysis of complex heart rate dynamics: discrimination of age and heart failure effects

Quantifying the complexity of physiologic time series has been of considerable interest. Several entropy-based measures have been proposed, although there is no straightforward correspondence between entropy and complexity. These traditional algorithms may generate misleading results because an increase in system entropy is not always associated with an increase in its complexity, and because the algorithms are based on single time scales. Recently, we introduced a new method, multiscale entropy (MSE) analysis, to calculate entropy over a wide range of scales. In this study, we sought to determine whether loss of complexity due to aging could be distinguished from that due to major cardiac pathology. We analyzed RR time series from young subjects (n = 26), elderly subjects (n = 46) and subjects with congestive heart failure (n = 43). The mean MSE measures of each of the three groups revealed characteristic curves, suggesting that they capture fundamental changes in the heart rate dynamics due to age and disease. We used Fisher's linear discriminant to evaluate the use of MSE features for classification. In discriminant tests on the training data, we found that MSE features could separate elderly, young and heart failure subjects with 92% accuracy and that older healthy subjects (mean age = 65.9) could be separated from subjects with heart failure (mean age = 55.5) with 94% accuracy. Also, we discriminated data from heart failure subjects and elderly healthy subjects with a positive predictivity of 76% and a specificity of 83% using only the MSE features. Larger databases will be needed to confirm if automatic classification results can match separation results. We conclude that MSE features capture differences in complexity due to aging and heart failure. These differences have implications for modeling neuroautonomic perturbations in health and disease.