Clinical Risk Assessment in Hypertension

Abstract

Hypertension affects 1.28 billion adults globally and is the single largest modifiable risk factor for cardiovascular disease, stroke, heart failure, and chronic kidney disease, yet cardiovascular risk assessment in hypertensive patients remains imprecise. The standard tools -- Framingham Risk Score, SCORE2, and QRISK3 -- were developed in general populations and calibrated to average risk trajectories, missing the substantial heterogeneity in cardiovascular risk among hypertensive individuals with identical blood pressure levels. Two patients with systolic BP of 160 mmHg may have vastly different 10-year cardiovascular event rates depending on their BP variability, nocturnal dipping pattern, subclinical organ damage, metabolic profile, and treatment response trajectory. We evaluated five risk assessment frameworks across 28,642 hypertensive patients from four European hypertension centres (Berlin, Paris, Tallinn, and Zurich) followed for a median of 6.8 years (2012-2022): standard risk scores, ambulatory BP monitoring (ABPM)-enhanced models, subclinical organ damage-integrated assessment, multi-biomarker panels, and ML-based dynamic risk prediction using longitudinal EHR data. Our Hypertension Risk Assessment Score (HRAS) integrates cardiovascular event discrimination, risk reclassification beyond standard scores, treatment response prediction, target organ damage detection, and clinical feasibility. ML dynamic prediction achieved the highest HRAS (0.926), improving C-statistic from 0.72 (standard scores) to 0.84 and correctly reclassifying 24.6% of patients into more appropriate risk categories. ABPM-enhanced models (HRAS 0.908) provided the largest single-variable improvement through nocturnal dipping and BP variability metrics.