A new study from investigators at Mass General Brigham shows that clues about hypertension may be buried in electronic health records (EHR). Using natural language processing, a form of artificial intelligence, researchers identified patients who had a heart ultrasound indicating thickening of the heart muscle, a condition frequently caused by hypertension. When physicians were notified of these results, they were almost four times as likely to diagnose hypertension and prescribe medications to control high blood pressure. This study highlights the potential for innovative, automated approaches that can use preexisting electronic health data to enhance treatment for patients with heart conditions. The results are published in JAMA Cardiology and were simultaneously presented at the 2025 American College of Cardiology’s Annual Scientific Session & Expo.
“Hypertension is known as the silent killer because people can have blood pressure that’s too high without having any symptoms from it,” said senior author Jason H. Wasfy, MD, MPhil, of the Cardiology Division, Department of Medicine at Massachusetts General Hospital (MGH), a founding member of the Mass General Brigham healthcare system. Wasfy is also a physician investigator at the Mongan Institute at MGH. “If they’re not getting checked for it enough, then high blood pressure can damage the heart and the vessels over time in a way that would have been preventable had the blood pressure been detected early.”
In the United States, nearly half of individuals with hypertension are unaware of or untreated for the disease.
“There is so much information that’s generated through routine clinical care, such as when you see your doctor or undergo a test. And there are often subtle clues in this information that may indicate a patient has hypertension. But it’s impossible for clinicians to master the entire medical record. The premise of our trial was that the data are likely hiding in plain sight, and we wanted to validate methods of bringing it to light to improve the care of our patients,” said lead author Adam Berman, MD, MPH, who conducted the study while in the Division of Cardiovascular Medicine at Brigham and Women’s Hospital, a founding member of the Mass General Brigham healthcare system. At the time, Berman was the David F. Torchiana Fellow in Health Policy and Management at the Massachusetts General Physicians Organization. Berman is now an assistant professor in the Department of Medicine, the Leon H. Charney Division of Cardiology at NYU Grossman School of Medicine.
The research team created and then used natural language processing that could sift through data from echocardiograms (heart ultrasounds) to identify cases of left ventricular hypertrophy, a thickening of the heart muscle often caused by hypertension. The algorithm identified 648 patients at Mass General Brigham who were not previously known to have any heart muscle problems and were not being treated for hypertension. The average patient age was 59 years and 38% were women. They randomized half of the patients to receive the intervention, and, for those patients, a population health coordinator notified the patient’s doctor of the finding. They also provided resources for additional care, including facilitating a 24-hour blood pressure monitoring test or scheduling an evaluation with a cardiologist. The clinicians for patients in the non-intervention control group were not contacted, and their patients were monitored under usual care.
Patients in the intervention group were nearly four times more likely to receive new hypertension diagnoses (15.6% vs 4.0%) and to be prescribed antihypertensive medication (16.3% vs 5.0%) than those in the control group. There was no meaningful difference in the number of follow-up appointments with primary care physicians between the groups. Clinicians mostly viewed the intervention positively — of the 82% who responded to the initial notification, qualitative scoring showed 72% had a positive reaction.
“There was a strong interest from our team in making sure this is something that physicians and patients would value,” Wasfy said. “Clinicians are often overloaded with alerts that can cause fatigue and burnout, so we intentionally designed our outreach to be delivered by a person.”
More work is needed to determine if this notification delivery method could be altered or automated for larger reach and easier implementation in other healthcare settings while maintaining effectiveness.
“The goal is to augment traditional care, using the data that already exist,” said Berman. “These patients have undergone testing, and their data are sitting there in a digital library gathering digital dust. Our trial demonstrates that we can harness these data to improve healthcare delivery and the treatment of our patients.”
Authorship: In addition to Wasfy and Berman, Mass General Brigham authors include Michael K. Hidrue, Curtis Ginder, Linnea Shirkey, Japneet Kwatra, Anna C. O’Kelly, Sean P. Murphy, Jennifer M. Searl Como, Yee-Ping Sun, William T. Curry, Marcela G. del Carmen, Ron Blankstein, David A. Morrow, Benjamin M. Scirica, Niteesh K. Choudhry, and James L. Januzzi. Additional authors include John A. Dodson and Danielle Daly.
Disclosures: A full list of disclosures can be found in the paper published in JAMA Cardiology.
Funding: The study intervention was funded by the Massachusetts General Physicians Organization in support of cardiovascular care delivery innovation.
