| A decrease in the variability in the time interval spanning cardiac contractions (as measured from R-complex to R-complex of the electrocardiogram (ECG)) has clinical significance for cardiovascular disease. Untransformed heart period variability (HPV) metrics, however, are often treated as minor variations on a similar theme and, as a result, both (i) the requirements of common HPV metrics from data, and (ii) the requirements of common data sets from HPV metrics remain misunderstood. This problem manifests itself in the prevalence of HPV papers in both medical and engineering peer-reviewed literature which use multiple metrics whose selection is unjustified and persists, in part, due to the absence of adequate quantitative assessments of the discrimination power of HPV metrics.;These results indicate, first, that low amplitude cardiac accelerations are important. The results indicate, second, that the treatment of CHF can benefit from the use pNNx but that either it must be combined with a second, complementary metric or placed at the output of a subset-isolating screening stage to improve efficiency.;The results of this work support the thesis that (a) a metric with the power to statistically discriminate individuals with congestive heart failure (CHF) from healthy individuals can be derived from RR interval data, (b) its discrimination power can be characterized consistently with results existing in literature, (c) the results in literature can be expanded with more powerful discriminant assessments, and (d) these results will generalize given a different sample of comparable data. Results include, first, the statistical formulation and derivation of a HPV metric. This measure can be taken as a measure of cardiac acceleration and possesses the well-known HPV metric termed pNNx as a subset. Second, p values in agreement with literature were obtained. Third, the physiological variation manifests itself most powerfully for pNNx parameter for x=16 msec. Third, at the cost of a high proportion of false positives, pNNx raised the visibility of the CHF sample (2% prevalence) by 11.4% and 34.1% while finding 95% and 85% of CHF subjects, respectively. A low rate of false positives (approximately 1 in 4) can be achieved if consideration is restricted to a CHF subset representing 15% of the total sample. Fourth, cross validation suggests these results are repeatable given a similar CHF sample. |
