Joseph Nano Artificial Intelligence, Cardiovascular Disease, Uncategorized

AI + ECG: A new way to flag future high blood pressure—years early

An algorithm trained on routine heart tracings predicted who would develop hypertension and who faced higher risks of heart and kidney disease. Here’s what that could mean for check‑ups.

By The Nano Post Health Desk

A large study in JAMA Cardiology reports that a new artificial‑intelligence tool can spot people at risk of developing high blood pressure (hypertension) just from a standard 12‑lead ECG—the same, quick test you get in many clinics and ERs. The model, called AIRE‑HTN, was trained on 1,163,401 ECGs from 189,539 patients and then tested in both U.S. hospital records and the UK Biobank research cohort. It predicted who went on to develop hypertension with solid accuracy (C‑index ~0.70) and added useful information on top of usual risk factors like age and current blood‑pressure readings.  

Why it matters

Hypertension is common, often silent, and a leading driver of heart attack, stroke, heart failure and chronic kidney disease. If clinicians could flag “quiet” risk during a routine ECG, they might counsel patients earlier on monitoring and prevention—before numbers climb. In this study, higher AIRE‑HTN scores were independently linked with later cardiovascular problems—even after accounting for traditional risks:

  • Heart attack: HR 3.13 per SD increase in score.
  • Heart failure: HR 2.60.
  • Cardiovascular death: HR 2.24.
  • Ischemic stroke: HR 1.23.
  • Chronic kidney disease: HR 1.89.  

How the AI worked

The system looks for subtle ECG patterns—tiny shifts in wave size, timing and shape—that humans may not notice. Signals tied to risk included differences in QRS voltage/duration, R‑wave progression, and T‑wave morphology, consistent with early, often invisible heart changes seen in rising blood pressure. Importantly, performance held up even in people with “normal” ECGs and without left‑ventricular hypertrophy, suggesting the model isn’t merely picking up obvious damage.  

Study at a glance

  • Where: Developed at a Boston academic center; validated in the UK Biobank.
  • Who: 19,423 U.S. outpatients (6.8 years’ follow‑up; 33% developed hypertension) and 35,806 UK participants (4 years’ follow‑up; 4% developed hypertension).
  • Performance: C‑index ~0.70 in both settings; improved risk prediction beyond standard clinical factors (continuous NRI 0.44 in the U.S.; 0.32 in the UK).  
  • Extra insight: People with high AI scores were also more likely to later receive blood‑pressure medicines in the UK cohort.  

What it 

doesn’t

 mean (yet)

This isn’t a diagnosis, nor a replacement for a cuff. The study used existing records and a volunteer cohort; it did not test real‑time clinical use or whether acting on AI scores improves outcomes. The authors call for prospective trials and careful rollout to ensure fairness across populations.  

What you can do now

  • Know your numbers: Regular, accurate home BP checks still rule.
  • Mind the basics: Healthy weight, salt‑smart eating, exercise, sleep and limiting alcohol all help keep BP in range.
  • Ask smart questions: If you’re getting an ECG (for chest pain, a check‑up, pre‑op, etc.), this line of research suggests ECGs might someday double as a BP‑risk screen.  

Bottom line: In one of the largest studies of its kind, an AI read of a routine ECG signaled who would develophypertension and who faced higher risks of major heart and kidney problems. It’s promising—and it underscores how familiar tests may soon pull double duty in prevention—but real‑world trials will need to show that using these predictions actually changes care and saves lives.  

Source: Sau A, Barker J, Pastika L, et al. “Artificial intelligence–enhanced electrocardiography for hypertension: Prediction of incident hypertension and risk stratification,” JAMA Cardiology*, 2025.*  

Editor’s note: This article is for information only and is not a substitute for professional medical advice.