Frontiers in Systems Biology · Nov 3, 2026
Development of a pulse wave velocity based decision tree algorithm for hypertension identification
By ÓSCAR MARTÍNEZ CIFUENTES, IRIS OTERO LUIS, ALICIA DEL SAZ LARA et al.
Background/Objectives: Current hypertension (HTN) identification relies primarily on conventional blood pressure measurements, which may fail to identify individuals with previously unrecognized HTN. Pulse wave velocity (PWv), an established marker of arterial stiffness, may provide complementary information on vascular status. This study aimed to develop a clinically applicable classification model derived from a decision tree model model that integrates systolic blood pressure (SBP), diastolic blood pressure (DBP), and PWv for the opportunistic identification of isolated HTN. Methods: A total of 769 765 adults were classified as "apparently healthy", "isolated HTN", or "HTN" on the basis of medical history, antihypertensive treatment, and blood pressure measurements. A supervised decision tree model incorporating SBP, DBP, and PWv was developed to derive a clinically applicable classification model. Model performance was evaluated using classification metrics, confusion matrices, and the area under the receiver operating characteristic curve (AUC). Sex specific models have also been developed to explore potential differences in classification thresholds. Results: DBP was identified as the root node of the decision tree, with an initial threshold of 77.75 mmHg, followed by SBP at a threshold of 130.25 mmHg. PWv contributed to classification at multiple subsequent nodes, with thresholds ranging from 7.35 to 10.65 m/s. Sequential integration of DBP, SBP, and PWv enabled discrimination among apparently healthy individuals and those with isolated HTN or established HTN. Sex specific models showed broadly similar hierarchical structures, although the optimal threshold differed between males and females. Conclusions: The integration of PWv with conventional blood pressure measurements may facilitate the opportunistic classification of isolated HTN. Although comparative analyses showed improved discrimination with PWv, an out of fold sensitivity analysis did not demonstrate a material improvement in macro ROC-AUC. Further studies are needed to establish its independent incremental value.