Research finds hidden hypertension clues in electronic health records

A new study from investigators from the popular general Brigham shows that cues about hypertension may be buried in the electronic health record (EHR). Using natural language processing, a form of artificial intelligence, researchers identified patients with cardiac ultrasound that exhibited cardiac muscle thickening. This is a condition that is frequently caused by high blood pressure. When doctors were notified of these results, they were almost four times more likely to diagnose hypertension and prescribe medication to control hypertension. This study highlights the potential for innovative and automated approaches that can enhance the treatment of heart disease using existing electronic health data. The results are published in Jama Cardiology At the same time, it was presented at the 2025 American Faculty of Cardiology Annual Science Session & Expo.

“Hypertension is known as a silent killer because people can have too high blood pressure without symptoms,” said Jason H. Wasphy Mphil, senior author of the Department of Heart Disease, Department of Heart Disease (MGH) at Massachusetts School of Medicine (MGH). Wadfy is also a physician investigator at MGH's Mongan Institute. “If they haven't checked it properly, hypertension can damage the heart and blood vessels over time in preventable ways if blood pressure is detected early.”

In the United States, almost half of people with high blood pressure are not aware of the disease or are not treated.

“There is a lot of information generated through routine clinical care, such as when you meet a doctor or take a test. There are often subtle clues in this information that may indicate that a patient has high blood pressure. However, it is impossible for clinicians to master the entire medical record. MD, MPH conducted research in the Department of Cardiovascular Medicine at Brigham and Women's Hospital. At the time, Berman was David F. Tortiana Fellow of Health Policy and Management for the General Physician Organization in Massachusetts. Berman is currently an assistant professor at the School of Medicine and the Leon H. Charney Circulation Department of NYU Grossmann School of Medicine.

The researchers sifted data from echocardiography (cardiac ultrasound) to identify cases of left ventricular hypertrophy, identify myocardial thickening that is often caused by hypertension, and created and used the natural language processing used. The algorithm identified 648 patients of the popular general Brigham who had not been treated for hypertension, previously unknown to have myocardial problems. The average age of patients was 59 years old, with 38% being female. They randomized half of the patients and received interventions, for those patients, the population health coordinator notified the patient's physician of their findings. They also provided resources for additional care, including promoting 24-hour blood pressure monitoring tests and scheduling assessments with cardiologists. Clinicians of patients in the non-intervention control group were not contacted and the patients were monitored under normal care.

Patients in the intervention group were almost four times more likely to receive a new hypertension diagnosis (15.6% vs. 4.0%) than those in the control group. There was no meaningful difference in the number of follow-up appointments with primary care physicians across groups. Clinicians were actively looking at 82% of interventions in response to the initial notification, so qualitative scoring showed that 72% responded positively.

There was a strong interest from the team in making sure this is something doctors and patients value. Clinicians are often overloaded with alerts that can cause fatigue and burnout, so we intentionally designed our outreach to be delivered by people. ”

Jason H. Wasphy, MPHIL, MPHIL, MPHIL, Medical School, Heart Disease Division, Massachusetts General Hospital, Medical Department, Massachusetts General Hospital

More work is needed to change or automate this notification delivery method to determine whether it can be automated for greater reach and easier implementation in other healthcare settings while still being effective.

“The goal is to increase traditional care using data that already exists,” Berman said. “These patients are undergoing tests and their data are sitting in a digital library that collects digital dust from our digital library. Our trials show that we can use these data to improve healthcare delivery and patient care.”