Artificial Intelligence Identifies Worsening Heart Failure

Researchers have developed a new wearable sensor that leverages artificial intelligence to identify worsening heart failure before a health crisis occurs, potentially preventing hospital readmissions.

Published in Circulation: Heart Failure, the study shows that the wearable device predicted critical changes in heart failure patients with a level of accuracy comparable to implantable sensors.

About 6.2 million Americans are living with heart failure, researchers said, and the condition is the top hospital discharge diagnosis in the US. Up to 30 percent of these patients will likely be readmitted to the hospital within 90 days of discharge, presenting with symptoms like shortness of breath, fluid buildup, and fatigue. For patients with heart failure, higher hospitalization rates often translate to worse outcomes.

“Those individuals who have repeated hospitalizations for heart failure have significantly higher mortality," said Biykem Bozkurt, MD, PhD, a study co-author, director of the Winters Center for Heart Failure Research at the Baylor College of Medicine in Houston.

"Even if patients survive, they have poor functional capacity, poor exercise tolerance and low quality of life after hospitalizations. This patch, this new diagnostic tool, could potentially help us prevent hospitalizations and decline in patient status."

The team noted that while implantable cardiac sensors have shown promise in reducing hospital readmissions for heart failure patients, the impact of wearable sensors hasn’t been tested. The group set out to develop a noninvasive device that could help identify worsening heart failure.

Researchers followed 100 patients with heart failure who were diagnosed and treated at four VA hospitals. After discharge, patients wore an adhesive sensor patch on their chests 24 hours a day for up to three months. The sensor monitored electrocardiogram (ECG) and motion of each subject.

The data was transmitted from the sensor via Bluetooth to a smartphone and then sent to an analytics platform on a secure server, which obtained heart rate, heart rhythm, respiratory rate, walking, sleep, and other activities. An AI algorithm determined a normal baseline for each patient. When data deviated from the baseline, the platform would trigger a clinical alert indicating that the patient’s condition was getting worse.

There were 35 unplanned non trauma hospitalization events, including 24 worsening heart failure events. The platform successfully detected critical changes in patients’ conditions before hospitalizations with 76 percent to 88 percent sensitivity and 85 percent specificity. On average, the prediction occurred 10.4 days before a hospital readmission took place.

"This study shows that we can accurately predict the likelihood of hospitalization for heart failure deterioration well before doctors and patients know that something is wrong," said the study's lead author, Josef Stehlik, MD, MPH, co-chief of the advanced heart failure program at U of U Health.

"Being able to readily detect changes in the heart sufficiently early will allow physicians to initiate prompt interventions that could prevent rehospitalization and stave off worsening heart failure," continued Stehlik, who also serves as medical director of the heart failure and heart transplant program at George E. Wahlen VA Medical Center in Salt Lake.

The wearable sensor could also be especially effective for heart failure patients who have recently been discharged from the hospital. The readmission rate in the first 90 days of discharge is approximately 30 percent, researchers noted.

“There's a high risk for readmission in the 90 days after initial discharge," Stehlik said. "If we can decrease this readmission rate through monitoring and early intervention, that's a big advance. We're hoping even in patients who might be readmitted that their stays are shorter, and the overall quality of their lives will be better with the help of this technology."

The team plans to further refine the sensor in future research.

“These results provide a rationale for the next step, a prospective study, currently in planning, which will randomize patients to an active arm—remote monitoring with alerts communicated to the clinical team and clinicians following a standardized response algorithm, versus control—remote monitoring without alerts being generated,” researchers said.

“This study should provide important insights into the clinical efficacy of wearable analytics in improving HF outcomes. A critical step will be implementation into clinical workflow and development of an algorithmic treatment response to system clinical alerts.”