Artificial intelligence driven clustering of blood pressure profiles reveals frailty in orthostatic hypertension.

Abstract

Gravity, an invisible but constant force , challenges the regulation of blood pressure when transitioning between postures. As physiological reserve diminishes with age, individuals grow more susceptible to such stressors over time, risking inadequate haemodynamic control observed in orthostatic hypotension. This prevalent condition is characterized by drops in blood pressure upon standing; however, the contrary phenomenon of blood pressure rises has recently piqued interest. Expanding on the currently undefined orthostatic hypertension, our study uses continuous non-invasive cardiovascular data to explore the full spectrum of blood pressure profiles and their associated frailty outcomes in community-dwelling older adults. Given the richness of non-invasive beat-to-beat data, artificial intelligence (AI) offers a solution to detect the subtle patterns within it. Applying machine learning to an existing dataset of community-based adults undergoing postural assessment, we identified three distinct clusters (iOHYPO, OHYPO and OHYPER) akin to initial and classic orthostatic hypotension and orthostatic hypertension, respectively. Notably, individuals in our OHYPER cluster exhibited indicators of frailty and sarcopenia, including slower gait speed and impaired balance. In contrast, the iOHYPO cluster, despite transient drops in blood pressure, reported fewer fallers and superior cognitive performance. Surprisingly, those with sustained blood pressure deficits outperformed those with sustained rises, showing greater independence and higher Fried frailty scores. Working towards more refined definitions, our research indicates that AI approaches can yield meaningful blood pressure morphologies from beat-to-beat data. Furthermore, our findings support orthostatic hypertension as a distinct clinical entity, with frailty implications suggesting that it is worthy of further investigation.