A machine learning score called HyperScore, developed from 566 imaging and clinical variables in 27,099 UK Biobank participants and validated in 5,507 ARIC participants, identified severe hypertension-related end-organ disease with an AUC of 0.964 and predicted incident organ-specific disease over 7 years. In the study, higher HyperScore stages were linked to worse survival and distinct multiorgan progression patterns affecting the heart, brain, kidneys, vasculature, lungs, liver, and metabolic systems.

Why It Matters To Your Practice

• Hypertension-related organ damage often develops before symptoms and can be hard to detect during routine care.

• This framework aims to quantify current multiorgan damage and estimate where a patient may be headed next, rather than relying only on blood pressure thresholds or traditional risk factors.

• If validated for clinical deployment, it could help identify patients who need more intensive monitoring or earlier intervention.

Clinical Implications

• HyperScore separated patients by severity of HTN-associated organ injury and was associated with survival differences.

• The model also generated organ-specific “HyperTrajectories,” suggesting that HTN may follow different damage patterns across patients.

• For clinicians, that raises the possibility of more personalized imaging follow-up, risk discussions, and phenotype-specific treatment strategies.

Insights

• The model used semisupervised contrastive trajectory inference on multimodal data spanning cardiac, brain, renal, vascular, pulmonary, hepatic, and metabolic domains.

• Performance remained consistent in external validation from ARIC, where trajectory progression predicted new end-organ disease over 7 years.

• The authors reported that HyperScore outperformed existing risk scores for predicting survival and incident multiorgan disease.

The Bottom Line

• AI may help clinicians move from diagnosing HTN by pressure alone to staging its cumulative organ effects.

• HyperScore is promising for risk stratification, but it remains a research tool until workflow integration, generalizability, and outcome impact are proven in prospective practice settings.