The prediction of mortality from continuous non-invasive cardiovascular signals on standing: entropy was significant, but not the overall response profile

In this study, a novel approach is presented using principal component analysis and sample entropy (SampEn) for the analysis of continuous blood pressure (BP) data measured non-invasively during an active stand (AS) in a large sample of older adults. The method allows for the extraction of the bulk trends from these data in the form of principal components (PCs), which can be used as independent predictors of outcomes, and greatly increases the stationarity of the remaining data, allowing for secondary analyses such as SampEn. The relationship between AS BP measures (SampEn and first 6 PCs) and risk of all-cause 8-year mortality was investigated via Cox proportional hazards regression models in a sample of community-dwelling older adults (
$\mathrm{n}=4873$
, with 209 deaths) from The Irish Longitudinal Study on Ageing (TILDA). Higher SampEn in BP signals was found to be a significant predictor of mortality risk. PC scores, which characterize the overall bulk changes in response to standing, were not significantly predictive of mortality when controlling for age, sex, and educational attainment. The quantification of signal entropy in continuously measured BP signals during AS could provide a clinically useful predictor of risk of death.